• State Resolved
  • Locked Locked
  • Replies 6 replies
  • Subscribers 85 subscribers
  • Views 307 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

F29H850TU: Regarding PWM Output Delay Time When Using EPWM_ICL

Kiyora Okuda
Prodigy 40 points

Part Number: F29H850TU
Other Parts Discussed in Thread: LAUNCHXL-F29H85X,

Hello experts,
I am currently evaluating EPWM_ICL on the LAUNCHXL-F29H85x evaluation board.

However, I have observed a delay time between PWM signals that pass through the ICL-X-BAR and those that do not. Furthermore, this delay time does not appear to be constant but seems to vary depending on the configuration. I predict that the delay time changes due to the ICL-X-BAR path. Could you please inform me of the maximum possible length of this delay time?

Thank you in advance!

  • 0
    TI__Mastermind 20380 points

    Hello,

    Yesterday was a holiday in the US, please expect a reply within the next day.

    Best,
    Matt

  • 0 in reply to Matt Kukucka
    TI__Mastermind 28580 points

    Hi Kiyora,

    I will need to reach out to our design team to get this information. In the meantime could you let me know the maximum variance of the delay time that you've been able to observe so far?

    Thank you,

    Luke

  • 0
    Prodigy 40 points

    Thank you very much for your reply.
    I’d like to share the details I was able to verify on my end.
    I tested with the following configuration:

    • EPWM1: 0°

    • EPWM1A: swapped with EPWM2A (via ICL-X-BAR) through ICL

    • EPWM1B: output as-is

    • EPWM2: 180°

    • EPWM2A: swapped with EPWM1A (via ICL-X-BAR) through ICL

    • EPWM2B: output as-is

    I checked the output waveforms on an oscilloscope, set the horizontal scale to 10.00 ns, read the divisions, and recorded how much the rising/falling edges were delayed relative to each other.

    No. Swap Signal 1 Signal 2 Slower Delay (ns)
    i No EPWM1A EPWM1B EPWM1A 2
    ii No EPWM2A EPWM2B EPWM2A 2
    iii Yes EPWM1A EPWM2B EPWM1A 4.5
    iv Yes EPWM2A EPWM1B EPWM2A 4

    If you notice anything I may have overlooked or have recommendations for additional checks, I would greatly appreciate your guidance.

  • 0 in reply to Kiyora Okuda
    TI__Mastermind 28580 points

    Hi Kiyora,

    I have shared this data with the design team, I am awaiting confirmation that your numbers are expected, I will let you know once I receive a reply.

    Thank you,

    Luke

  • 0 in reply to Luke Jones
    TI__Genius 10351 points

    The experts are OOO this week, please expect a delay in response.

    Thanks

    Aswin

  • 0 in reply to Aswin P
    TI__Mastermind 36323 points

    Hi Kiyora,

    Apologies for the delayed response here! I will be picking up the thread from Luke.

    Knowledge Gap — No ICL-X-BAR-Specific Delay Spec

    I want to be upfront: the TI documentation (TRM and datasheet) describes the ICL-X-BAR architecture and its position in the signal path, but does not publish a dedicated propagation delay specification for the ICL-X-BAR itself [1]. So the answer below is based on the closest applicable datasheet specification and your own measurements.

    Your Measurements Are Consistent With Expected Behavior

    The delay you're seeing is real and expected. The ICL-X-BAR sits in the ePWM signal path after the Dead-band/Diode Emulation submodules and before the final output pins [2]. Signals routed through it traverse additional crossbar and multiplexer logic, which introduces propagation delay compared to signals that bypass it [1][3].

    Your results align well with this:

    Case
    Path
    Your Measured Delay
    No swap (A vs B, same module)
    Signal A passes through ICL logic; B does not
    ~2 ns
    Swap (A from other module vs B)
    Signal A passes through ICL + ICL-X-BAR crossbar
    4–4.5 ns

    Maximum Delay Bound

    The F29H850TU datasheet specifies the maximum skew between any two ePWM outputs as [2]:

    Path
    Max Skew
    Shortest path
    5.1 ns
    Longest path
    8.9 ns

    These specs apply to all ePWM outputs regardless of whether HRPWM is used [2]. Since the ICL-X-BAR is part of the ePWM output path, the 8.9 ns longest-path skew specification represents the worst-case upper bound for the delay difference between any two ePWM outputs, including those routed through the ICL-X-BAR versus direct paths.

    Your measured 4–4.5 ns is well within this 8.9 ns maximum.

    Why the Delay Varies With Configuration

    The variance you observed is expected because the XBAR inputs are organized into multiplexer groups [3], and different swap configurations route signals through different path lengths within the crossbar. The specific group and input selection determines the propagation delay for that particular configuration.

    Summary

    • Your measurements (2–4.5 ns) are within expected range.
    • The datasheet-specified worst-case bound is 8.9 ns (longest-path ePWM output skew) [2].
    • There is no separate, published ICL-X-BAR-specific delay figure — the 8.9 ns overall ePWM skew spec is the best available upper bound.
    • If your application requires tighter timing guarantees specifically for ICL-X-BAR paths, I'd recommend filing another ticket to specifically request the TI design team for silicon-level characterization data.
    1. F29H85x Technical Reference Manual — XBAR and ICL Architecture
    2. F29H850TU Datasheet — ePWM Switching Characteristics (Skew)
    3. F29H85x XBAR Application Guide

    Best Regards,

    Zackary Fleenor