Page tree

Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

Dashboard Subsystem (Adi & Wenstrup 6/5/19):

  • Things we liked this year:
    • Everything was custom, but not much over engineering
    • simple and effective design
    • Barring CAN on steering, everything worked!
    • Wrote code for drive modes (still need to implement)
    • Debugging LEDs were godsend
    • Lots of learning to be had
    • Good opportunities for interaction with mechanical
    • SS LEDs were very helpful
  • Things we want to implement next year:
    • Drive modes
    • Get rid of unused systems
      • Op amps on throttle board
    • Quick respin to get rid of any rework done last minute
    • All boards use the same buck converter
    • Add an RTD light
  • Things we feel need to change next year:
    • LEDs for IMD and AMS
    • Get the buck converter off the steering board... ??????????? why is it even there
    • Go back to black enclosures 
  • Stretch goals for the coming year:
    • Have a large LED matrix in place LED bars
      • Its the same adafruit library 
      • Will look cooler and we can shove more info in
      • Potentially numbers
    • Consolidate both boards into one boards
  • Random Ideas:
    • Dashboard overall is kinda ugly - we can try to pretty it up with aesthetics 
    • Bucks on Interface boards
    • 5V and Ground bus bars in front of car, to minimize board I/O
      • judges are anal about fuse protection
    • Add functionality to reduce throttle at low battery (maybe with driver input as well)
    • Sand the back of the steering wheel
    • Plan around tasks that are often completed last minute, like harnessing.
      • It becomes such a time sink  at the end of the year
      • It wouldn't have to be if we had more structured first year training on it
        • It is a task where more hands make a huge difference. 

Steering (Adi & Manu 6/5/19):

  • Things we liked this year:
    • Openness is good. Freedom to explore
    • Added electronics to the steering wheel!
  • Things we want to implement next year:
    • I want to get the CAN connection working.
    • Explore new connectors/connections
      • Encourage them to explore and learn
    • Send it 5V instead
  • Things we feel need to change next year:
    • Despite the openness, be sure to provide ample guidance
    • Have a specific goal in mind
  • Stretch goals for the coming year:
    • The following would require complete redesign
    • LCD
    • Arrow to show optimal acceleration and driving. (Not necessarily a one year goal)
  • Random Ideas:
    • More Buttons
    • More LEDs
    • Sand the steering wheel

LV Interface Subsystem (Adi & Manu 6/5/19):

  • Things we liked this year:
    • Box design was good.
  • Things we want to implement next year:
    • LCD display
    • Consider a hinge design for the lid
  • Things we feel need to change next year:
    • Would love to see next iteration.
    • Double Fault LEDs
    • Full Debug Box.
    • Add all necessary + 5 extra debugging LEDs
    • Make sure the master panel is mechanically correct before waterjet
    • No jank stuff on boards before competition 
  • Stretch goals for the coming year:
    • Treat like a legit project. Don't just put it aside
  • Random Ideas:

Random Ideas for Next Year (Adi, Wenstrup, Manu 6/5/19):

  • Buy and use 22 gauge wire for everything that's not power 
  • Consider keeping board IO so we can reuse MKIV boards
  • Put motor controller on wiring harness
  • Explore new wiring/harnessing methods
    • Tape
    • Shielding
  • Read rules for HV wiring strength; plan for and buy acceptable conduit
  • Include applicable E tech rules in all meetings, design reviews, confluence pages, and wherever else
    • Make sure all are accounted for !
  • Turn the LV box into a legit electrical and mechanical project
    • Have LED indicators on the outside
      • IMD, AMS, RTD, Debugging, etc
  • Have a wire station
    • Hang wires so that theyre easy to cut, find, and access
    • Hang jumper cables, scope wires, etc
  • Have 2 populated and working copies of every board
  • Consistent way of dealing with wires
    • Create a standard wiring system, colors, diagram
  • Clean up programming box

Talk with Anil (Adi, Anil 6/6/19):

  • Things we liked:
    • Every first year had a project
    • Transfer of information on BMS
    • Enthusiasm for the accumulator this coming year
  • Things we want to change:
    • First years should not be solely responsible for service section projects
    • Utilize container space more effectively
    • Haul ass on sponsorship - dont focus on the same companies, contact everyone
  • Things we feel need to change:
    • Implement sensing 
      • Get sensors spec'd out in the first couple of months
    • Electrical and mechanical working closer earlier 
  • Concerns for new Electrical Team:
    • Who will committedly work on sensing from the start
    • Who will lead the service section, probably Adi
    • Is the transfer of information going to be consistent bc of the more decentralized nature of the team

BMS Talk (Adi & Hoppe 6/6/19):

Please Note: These notes were taken very fast by someone, me (Adi), who doesn't know too much about the BMS

Excuse any errors and feel free to fix them.

  1. The LV startup sequence worked fine
  2. When we enabled the rinehart, the issue began
    • Throttle sent inverter enable when RTD
    • Resulted in TS shutdown
  3. There are a couple of things that can cause this:
    • Depends on at what point the AMS fault is tripped
      • 2, 6, 10 seconds
  4. According to Corey (6/8/19): 
    • AMS fault happened as soon as the RTD finished
    • IMD fault turned on immediately after we closed TS
  5. Pierre (E tech guy) didn't seem to think this was a huge issue. He suggested:
    • Enabling the inverter causes lots of noise (10 kHz) and a harmonic at 1 MHz/10 MHz
      • That harmonic showed up and backpropogated in the entire battery pack
        • Showing up on the voltage sense lines
        • Could be LC coupling into the isoSPI lines
    • Lots of vulnerability to the noise:
      • Starting the inverter could have caused us to start losing BMS messages
        • If we had bit errors, the BMS 15 bit error protocol (verified by CRC) will be seen in CRC or data
  6. Our measurement cycle is 2 seconds long
    • Kind of slow
      • A limitation of UART
    • At tech inspection, all they want to see are the values
      • They trust that you have a well made system
    • There's a voltage read and a successive temperature read
      • If any of those throw a package read fault, it will throw out the data (missed cycle)
      • If there are 3 missed cycles in a row, the shutdown circuit will open
        • Usually works fine, isoSPI are noise robust, high traffic is not an issue
        • A single bit error here and there is not an issue and will be thrown out. Low error rate
      • It's not unlikely that turning on the inverter caused us to start losing packets
  7. What can we do?
    • Code that needs to be written:
      • Quantify how many packet read errors
        • 6 isoSPI messages per voltage + a lot (80-100 isoSPI messages for temp)
          • If any of them throws a bit error, the whole cycle is thrown out
      • Instead of a flag, count the number of issues during a measurement cycle.
        • If its small, we can change the code - the error checking algorithm
        • If its like 100 (big) errors, something drastic has to be done for noise reduction
        • If its random, then its not a problem, we can allow 1 or 2 bit errors per cycle
          • If its the same message, figure something else out
    • We must confirm the issue - is enabling the Rinehart what's causing the traffic?
      • How much noise?
        • If its small, we can write a more lenient algorithm that is still safe but handles errors better
          • Keeps the valid data?
      • If everything is getting lots of noise all the time, look for a hardware fix
        • Try testing a shielding and soldering a lead to it and connecting it to low side of cells
          • Try remaking the isoSPI wires
            • The pigtails are long, unsure of effects
            • Connecting the leads across the shield of the isoSPI would be tricky
              • would be metal and conductive
              • Be wary of the transformers
              • Basically, find a way to make isoSPI wires better
      • If not isoSPI wires, change all the voltage sense leads
        • There's an RC low pass filter - change the damping coefficient
      • A possibility is that we used the wrong dielectric cap (not the same on all boards)
        • The bias derating could be causing a resonant frequency that we are not expecting
          • the resultant could make an effectively lower resultant capacitance
            • Ask Lucky about DCD rating
    • Worst case: redo the BMS
      • Very time consuming and expensive
        • Spin a new BMS with a TI battery management chip
        • Reach out to CMU, get the application note and read the big description about filtering/BMS stuff
        • Make a nice LRC filter at 40 dB instead of 20 dB
        • Add a Zener Diode to clamp voltage in case a cell drops out (DO THIS NEXT YEAR EITHER WAY)
    • With HV gloves (SAFELY) AC couple scope probes and check across TS+ and TS-
      • Make sure everything we use has a 1kV rating before trying this
        • Look for the switching noise coming back into the bus
          • What does it look like?
        • Assume whatever AC signal we see there is the same as whats going on in the V sense lines
    • One mitigation we can have in code (unsure why I wrote this specific line):
      • If the noise frequency is similar to the communication frequency, that would be a problem
        • Last years metal box might have attenuated the noise a bit
        • isoSPI is 1 MHz, might be close to a strong harmonic from the rinehart noise
          • Consider changing to 250 kHz - bit timing be less affected
        • In the past, it was wire, not PCB trace
          • The impedance would actually be higher for PCB traces so probably not a problem
        • Frequency was 4 times slower last year
          • Try this first! Divide by 4 and see if it improves it
          • If it works, just run with it
  8. Reproduce the problem
  9. Try to fix it in code
    1. Different frequency
    2. New noise algorithm
    3. New isoSPI wires
  10. Respin boards
  • Other Notes:
    • We never ran into issues with rinehart making noise in the past
    • Might be due to higher impedance from cells this year
    • Amplified by number of cells and other stuff
    • Lithium Ion cell voltage across them sagged very little when under load
      • Might be part of problems, chemistry was fast enough to counter switching noise
      • Impedance was low enough at the high frequency that it (signal) just absorbed it
    • Pierre kept mentioning voltage sense leads - capacitively straight through the sense leads
      • A piece of copper (bus bar?) through the entire BMS
    • Consider exploring impedance from magnetic field tricky
      • Maybe try a big piece of sheet metal
        • Drill 2 small holes for leads, ground well and connect to chassis
    • Next year, regardless, include systems for lots more EMI filtering, be more deliberate with isoSPI wires
    • If we are experiencing magnetic interference, laminate some magnetic material on the outside of the garolite 
      • Join it well, so we have a good way to ground stuff on accumulator box and will do the job of shielding 


Advice from Hoppe (Adi & Hoppe 6/6/19):

  • Fix the BSPD short detection tests
    • We are experiencing weird frequencies
  • Do even less than what feels like too little
  • Create a bill of requirements 
    • Close the loop, follow up, be more thorough
    • Go through the hard requirements with people not making the circuitry
  • Something that would help on the interface sheet and bench top harness 
    • Name the nets better, will properly compile in KiCad 
    • Explore net ties on interface boards
    • Make names super consistent 
      • Struggled with shutdown in past - whats coming in and out? What direction?

Talk with Corey (Adi & Corey 6/8/19):

  • Things we liked:
    • First year retention was high, especially in the last 2 months
      • First years were capable at the end of the year
    • Boards were shipped in a timely manner
    • LPB was decently organized most of the time
    • Rather than 1 person who knows everything, there were several who knew what they were doing
    • Not much of a bro culture on electrical
    • People showed up to workdays at the end of the year
    • Standups at the meetings
    • People showing up on time (sometimes)
    • Checkins DO THEM PLEASE
  • Things we want to change:
    • Try not to double up people on projects
      • Was told this was a bad idea.
      • Only one person was left, every single time
    • Not redesigning every board on the car
      • Don't redo work that's done
    • Entrance interview: Don't ask about prior experience 
      • Doesn't matter. Didn't effect outcome AT ALL
    • Be on top of shipments (more)
  • Things we feel need to change next year:
    • Seating arrangements 
    • People HAVE to use confluence
      • Ask people at check-ins
      • As people at stand-ups
    • Don't waste resources
    • Standardize all connectors
  • Stretch Goals for the coming year:
    • LCD Screen
    • Sensors (DO THIS)
  • Random Ideas:

Advice from Corey (Adi & Corey 6/8/19):

  • Just because a confluence page is made, doesn't mean people will use it.
  • Go in with the mentality that kickoff meetings might not work
    • Intimidating for the new members (surrounded by knowledgeable upperclassmen
  • The job of a lead is to motivate their teammates
    • Make people see you as a person/friend rather than figure
  • Look into plug and play solution for sensing for Mk4 & Mk5
    • dont reinvent the wheel
    • explore ADC to CAN solutions
  • Try not to talk about how much work there is for formula all the time
  • Keep info all in one place
  • 1) Understand Mk4
    • You cannot improve on it if you dont understand it
    • Redo the look
    • Really understand the block diagram
    • Understand each board
    • Make a list of changes we want to make
  • 2) Read the rulebook - everything past IC
  • 3) Look at Corey's confluence page
    • Make your own confluence page
  • 4) Read the ESF
  • 5) Ask Questions
  • Start the ESF thanksgiving break at the latest
    • Give people the time to familiarize themselves with the system
    • Delegate ESF tasks
  • Start FMEA soon, not too much should change
  • You're always going to have 20 things you want to do
    • You can only do 10
  • Make a list of things that need to get done this summer

Advice from Corey (Adi & Corey 6/8/19 later in the day):

  • Last summer spent most of the time studying car
    • Make block diagram
    • Make interface sheet
  • Spent a lot of time thinking about onboarding
    • Project that people will take on
      • First ask leadership, then talk to returning members, and so forth
  • Show up before the meeting to plan out and make sure meeting runs smoothly
  • Answer questions during the meetings, your work outside of the meetings
  • People will quit if they don't think their project is important or if they don't know its role
  • Plan to have people banging their head for a month
  • Get people comfortable asking "dumb questions"
  • Don't assign projects you don't fully understand
  • Tutorials
    • Fast paced, but okay
  • Filtering for the super motivated people is not sustainable
    • Spent time this year trying to lower the barrier of entry
    • ACTIONABLE: Figure out what the barrier for time and passion is
  • 1-on-1 time is the most valuable time with members
    • You have to hand hold sometimes
  • When/How to cut people off. Set the deadlines
  • You will get frustrated walking people through something. Deal with it
  • People will drop as time goes on
    • Roll with the punches
    • Hand off projects popping up to people with the bandwidth
    • If that's not an option, do it yourself 
  • Please take 12 credits
  • Fuck the ESF (not actually, its a lot of work, plan accordingly)
  • ESF and FMEA will take a lot of energy
  • Schedule started to fall apart when the car wasn't done by spring break
  • ENSURE that code works on the harness. Check your assumptions. Check everyone's work
  • Resolving issues with working with people
    • Be ready to have a hard talk with people
  • Read the Rules
    • Read the rules
    • Read the Rules
    • Read the rules
  • Make the design decisions yourself, use upperclassmen as a resource
  • Don't make any assumptions, ask if you're unsure about a design decision
  • Big thing: Be clear about your expectations and understand other people's expectations
    • Extremely clear
  • Enclosure design reviews are useful - especially second semester
  • You can't force people to do work
    • Put people in a position to get their work done
    • If they aren't following through, have the expectations talk
      • Standups are great for putting people on the spot
  • Harnessing:
    • Wiring might getting fucked up because of changes of physical car
    • If you standardize connectors pinouts, make its explicit 
  • Putting datasheets on schematics is really useful
  • Interface boards should have silk screen all over them
  • Review every single decision before shipping boards
    • It will end up with a big waste of time for the small time to put in
  • Personally review every single review and layout that gets finalized
  • People will make mistakes
  • Be wary of decentralized onboarding
  • No one will read a list, except Corey (feels strongly about this)
  • Don't waste people's time - be mindful
  • Read sensor inputs from opamps - we dont want to pull current 

IMU & Sensing (Adi's Notes 6/2/19):

  • Things we liked this year:
    • Zero to a close to functioning IMU 
    • Messed around with lots of sensors
  • Things we want to change next year:
    • Better organization
    • general coherience of a sensing subteam 
  • Things we feel need to change next year:
    • Sensing needs to happen
  • Random Ideas:
  • IMU:
    • Function: Accelerometer, Gyroscope, 6 Axis Sensor
    • Inputs: BMI 088 (current sensor being used) 
    • Outputs: movements and rotations in 6 degrees of freedom; I2C or SPI 
    • Left to implement: Code
    • Needs to be positioned very close to COM

Post North Electrical Discussion (Adi's Notes 6/2/19):

Members present: Hoppe, Lucky, Corey, Adi, Wenstrup, Manu, Vienna 

  • Spend less time debugging before integrating 
    • We spent a good 5 weeks just debugging code on the harness
    • Ran into integration issues down the line that wasted precious time
  • Bang out the roadkill harness in 1 weekend
    • Set a required time block for electrical
  • Upperclassmen lead PCB projects
    • No need for redesign, its a big waste of time - 1 year 
    • First years help with testing and new implementations
    • Sophomores get to continue where they left off
  • Make MkV the next iteration of MkIV
  • Make electrical subteams:
    • Cockpit - subteam lead: Wenstrup - Dash R, Throttle, Steering
    • LV Box - subteam lead: Manu - LV interface, BSPD, Brakelight, CAN logger, LCD interface, etc.
    • Service Section - subteam lead: TBD - AIR control, TSAL, AIL, IMD Latch
    • Sensing - subteam lead: TBD - IMU, Wheel Speed, etc.
  • Have subsystem design reviews instead of subteam design reviews
    • Ex: Dashboard DR, etc.
  • Make a collective confluence that all parts of the subsystem update constantly
    • Ex: Dashboard Confluence: Updates to PCB, new Layouts, updates to enclosures, fabrication progress, etc
  • Kickoff meetings > Check-in Meetings > Critical Design Reviews:
    • all members of a subsystem have a meeting
    • Kickoff - super high level, list all rules, requirements, goals, etc. Block diagram level, understand how groups work together
    • Check-in updates of fab, progress, etc.
    • Critical DR - extremely detailed, everyone should leave with full knowledge of their system
  • Move check-ins to Saturday morning
    • Incentive to make people show up for workdays
    • Can still have frequent check ins without cutting into the 4 hours of meeting times
  • Make more people excited with integration
    • Work closely with mechanical
    • start the roadkill harness early in the year