A bit bitter experience with the Jetson Nano devices

I would like just to inform everyone about the hurdles and inconveniences I had to encounter when I tried to use the Jetson Nano A02 (4GB) and B01 (2GB) as a brain of a constructed device. I hope it saves precious time and nerves of the users. I hope also the engineers working on the Nano would fix the problems below in the nearest future…

  1. Power regulation system is inefficient.
    1.1 When I supply the power either with the GPIO pins 2 & 4 or the external socket, I need to deliver at least 5.5 V or the “overcurrent throttling” occurs, even when the JNANO is idle. The specs tells 5V is enough… The issue seems not to be present when powering the device from the USB3
    1.2 Speaking of USB… using two separate USB (micro USB 2 and 3) sockets for uploading the firmware and powering the device looks totally redundant and inconvenient. It also takes space on the carrier board. Couldn’t the microusb 3 be used for both purposes?
    1.2 If I want to drive a motor with JNANO PWM GPIO, I have two options: either to provide a JNANO power source that is totally separate from the motor power source or use extra buffer batteries. Otherwise a brown-out reboot occurs due to a too high voltage drop when the motor is turned on. A supercapacitor protecting the power from sudden drops, would be very helpful. I know at least one robot vendor that has resigned to use the Jetson devices exactly due to this problem.
  2. GPIO issues
    2.1 The TXB1008RYGR level converter sucks with its current output capabilities measured in microamperes ! Thus, if you want to use any GPIO line, an external buffer is a must. And it must be placed as close to the GPIO socket as possible, which complicates the design of a GPIO connected devices, a lot.
    2.2 The TXB1008RYGR causes issues with the I2S bus, rendering it virtually unusable (there was a post about that earlier by someone else, I had to deal with that too). I ended up using the I2S taken from the M2 slot.
    2.3 Devoting at least a single core for real-time GPIO handling would be really nice. Right now, using the device for real-time controlling the peripherials is barely possible without a substantial effort.
  3. The I2C
    3.1 Bricking the Jetson by a careless usage of the I2C bus is very easy, by overwriting its vital EEPROM data. It happens pretty frequently to a new users, it happened to me as well. Could the NVIDIA engineers provide an easy way to restore the EEPROM contents during boot or via a PC connected to the USB?
  4. Slow software update cycle.
    4.1 Well, ubuntu 18 is a bit outdated, isn’t it ?

I am sorry if I am complaining so much… but with the current prices, I really expected a bit more than what I currently get with Jetson devices.