Kira Adapter


Open questions:

  • Ethernet ("1000-BaseFX") on GTR or GTH?
  • What clock(s) are needed to support this?
  • What about other PL clock(s)? (none on the Kria except a PS clock)

Si5344 are unobtainable. Best use some fixed-freq LVDS oscillators.

  • GTR refclk for Ethernet, etc. Is a single 100MHz OK?


Power: Header with 5V, 3.3V 1.8V inputs.

LEDs on all supplies.

DC/DC modules are hard to source. Probably need to use an LTM4622.


Looking at power. SM/Adapter must provide:

  • The main 5V for the SoM (DC/DC from 12V)
  • VCCO for HPIO and HDIO
    • HPIO are 1.8V
    • HDIO are 3.3V

Power-up sequence:

  • Carrier deasserts POWER_OFF_C2M_L (5V pull-up, drive with open-collector or MOSFET from ZYNQ_EN).
  • SOM asserts VCCOEN_PS_M2C
  • Carrier turns on PS peripheral power (PS MIO and PS GTR connected devices)
  • SOM asserts VCCOEN_PL_M2C
  • Carrier turns on PL power (GPIO connected devices and VCCO to SoM)

Power inputs on SOM (from pinout)

  • VCC_SOM (5.0V max 4A)
  • VCC_BATT (1.2-1.5V battery or GND)
  • VCCO_HPA (Bank 66) 1.8V max 1A
  • VCCO_HPB (Bank 65) 1.8V max 1A
  • VCCO_HPC (Bank 64) 1.8V max 1A
  • VCCO_HDA (Bank 45) 3.3V max 1A
  • VCCO_HDB (Bank 43) 3.3V max 1A
  • VCCO_HDC (Bank 44) 3.3V max 1A

On the SM we have, controlled by ZYNQ_CORE_GOOD (Merc A-12 "PWR_GOOD")

  • +3V3_Zynq (max 3A)
  • +1V8_Zynq (max 3A)

Hopefully we can just use these directly for VCCO_HPx and VCCO_HDx. Important power-related status/control signals on Kria:

  • VCCOEN_PS_M2C - permission to power up MIO peripherals
  • VCCOEN_PL_M2C - permission to power up GPIO peripherals and VCCO
  • PWROFF_C2M_L - master power-up for SoM

Optional power-related stats signals (connect to IPMC or at least header/LEDs?)

  • PWRGD_FPT_M2C - power good for "full power domain" rails
  • PWRGD_LPD_M2C - power good for "low power domain" rails
  • PWRGD_PL_M2C - power good for all PL power rails
  • PS_ERROR_OUT_M2C - PS error indications
  • PS_ERROR_STATUS_M2C - PS error status (?)

So in the end I think we can route:

  • ZYNQ_EN to POWER_OFF_C2M_L with OC driver

Need to check levels on VCCO_EN_PS_M2C. On the SM this is the "Run" input to LTM4622 so should be fine (~ 1.2V threshold).


Enclustra uses total 64 3.3V IOs while Kria has only 48 available.

Connector A: 17/64 are undefined direction (I2C, GPIO), so should be on Kria 3.3V banks.

  • 17+23 = 40 on 3.3V bank (leaving 8 unused now)
  • 24 on 1.8V bank with fixed-direction level shifters (maybe 6 NLSV4T244)

Connector B: 48 1.8V I/Os used, with 84 HP I/Os available on Kria

Wirelist codes

  • 3VI - 3.3V input (carrier to module). Use level shifter
  • 3VO - 3.3V output (module to carrier). Use level shifter
  • HD - connect to HD I/O (1.8V)


Made a new big spreadsheet (mostly yesterday) with all the pins.

Dan says:

  • I2C for SI chips should be on the PS
  • UART for console should be on the PS
  • Other I2C and UARTs can be on the PL

UARTs are mux'd as follows:

  • "console" is either front-panel or IPMC controlled by USB_UART_EN. Then 4-way mux as:
    • Zynq PS UART
    • Debug header J701
    • Mezz 1
    • Mezz 2

All are 3.3V. We'll need a level shifter.


Power for Kria is 5V but SM delivers 12V. Need a hefty DC/DC.

Enclustra has e.g. XCZU7EV-2FBV900 (we ordered ME-XU8-7EV-2I-D1 on 3/27/20)


Ethernet on Rev 2a:

  • ESM has 6 ports
    • SGMII 0 - front-panel RJ-45 via PHY on SM
    • SGMII 1 - not used
    • PHY 0 - Zynq port 1
    • PHY 1 - ATCA A
    • PHY 2 - ATCA B
    • PHY 3 - IPMC
  • Enclustra has 2 ports wired:
    • PHY 0 - rear mag jack
    • PHY 1 to ESM PHY 0

So on the adapter ideally there would be two PHY, but DP83867 are unobtainable (72 weeks lead time!). If we just want to test all the GPIO etc we could operating without a PHY to the adapter for some time. We do need Ethernet to the Kria though. What about an SFP?

Last modified 17 months ago Last modified on Jul 14, 2022, 9:11:57 AM