• State TI Thinks Resolved
  • Replies 2 replies
  • Answers 2 answers
  • Subscribers 64 subscribers
  • Views 63 views
  • Users 0 members are here
Support feedback
Options
Options
Related

TPSM82823: Power Sequencing Issue with TPSM82823SILR on FPGA Board

Diwakar R
Prodigy 10 points
Part Number: TPSM82823
Other Parts Discussed in Thread: , TPSM82810

Tool/software:

We are using two boards PSB(power supply board) and SPB (Signal processing Board) in our design:

  1. PSB (Power Supply Board):

    • Generates FPGA power rails.

    • 8V → 5V using LTC3119EUFD#PBF buck converter.

    • 5V → 0.85V, 1.8V, 1.2V, 1.1V, and 3.3V using multiple TPSM82823SILR modules.

    • Sequencing is handled using PG and EN signals.

    • Provides power to the SPB through a mating connector.

  2. SPB (System Board):

    • Contains a Xilinx MPSoC FPGA (Zynq UltraScale+ 1CG series).

    • Receives all rails from PSB.

Issue:

  • When we test the PSB standalone, all rails power up correctly.

  • When we connect the SPB to the PSB before powering on, only the 0.85V rail comes up. Other rails like 1.8V and 1.2V do not turn on.

  • If we first power on the PSB and then hot-plug the SPB, all rails (0.85V, 1.8V, 1.2V, etc.) come up as expected. 

    Request: Could you help us understand why the TPSM82823SILR modules do not enable the rails when the SPB is connected at power-up?

  • 0
    Intellectual 330 points
    • Diagnose: Use an oscilloscope to monitor output voltages, 5V input, EN/PG signals, and current during power-up. Look for sawtooth waveforms indicating hiccup on failed rails.
    • Mitigate:
      • Reduce SPB capacitance on affected rails (e.g., minimize bulk/decoupling caps if possible).
      • Add load switches (e.g., TPS229xx) on PSB outputs to pre-charge or isolate SPB during startup.
      • Ensure PG pullup resistors (1–10 kΩ to 5V or independent source) are strong enough; avoid floating pins.
      • Test with dummy resistive loads on PSB to simulate SPB inrush and isolate if it's capacitance vs. FPGA leakage.
      • If feasible, use TPSM82823A variant for forced PWM (better under light/pre-biased loads) or higher-current modules.
  • 0 in reply to SUNILBHAI PATEL
    TI__Mastermind 21560 points

    Hi Diwakar,

    There is too much output capacitance on the rails. TPSM82823 supports only up to 47uF output capacitance. The TPSM82823 is probably going into current limit when trying to startup with such high output caps. The internal compensation is also not designed for such high output caps. 

    I would recommend to use TPSM82810 as it can support more output caps and also has adjustable softstart time. You need to have a larger softstart time to prevent the device from going into current limit while charging large output caps.

    Best regards,

    Varun