• State Resolved
  • Locked Locked
  • Replies 14 replies
  • Subscribers 83 subscribers
  • Views 470 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.

Original question:

TMS320F28379D: QEP output emulator using CLB logic

TMS320F28379D: Emulating a QEP signal on C2000

Wil P.
Expert 1660 points
Part Number: TMS320F28379D

I want to precisely emulate a QEP sensor ABZ (or ABI) differential output using 28379D and would like this signal to externally feedback into a QEP conditioning circuit and ultimately back to the MCU. I would like the ability to change the output frequency during run-time and I want the index (Z, from now on) to be aligned with A exactly once per rotation. I believe using three ePWM modules (6 signals total) would be the easiest, but I am open to suggestions.

To generate A and B, I configure them as UpDown counters, link TBPRD and CMPA values and then add a TBPRD/2 phase delay to B. Whenever an output frequency change is desired, TBPRD, CMPA and TBPHS(B) are all updated.

I am finding the biggest challenge to be properly aligning the index when the PPR of the QEP is large (for now, I am using 512, but I'd like a flexible solution).  Here are a few things that I have tinkered with, and some challenges that I have had:

QEP Position Interrupts (Janky McJankyPants)

  1. Configure the PWM module for Z identically to A and ensure shadow-loaded AQs.
  2. Set the output of Z disabled via AQ.
  3. Configure the QEP Position compare interrupt to trigger on the PPR-1 PWM period.
  4. In the interrupt, enable the output of Z using AQs and set the position compare to 0, so that it triggers immediately in the next cycle.
  5. In the next interrupt, go to 2.

Challenges here:

  1. Adds a new interrupt and QEP config that will not be in experimental control code (even if on separate core).
  2. Interrupt frequency can be high at low PPR and/or high frequency outputs.

TBPRD Stretching (Queen of the Janks)

  1. Configure the PWM module so that the clock divider for Z is PPR - (assume the dividers for A and B are each 1).  For PPR=512; ClkDiv=128 and HspClk=4.
  2. Configure Z as a up (or down) counter and set the AQ so that it gets turned on at the very end (when aligned with A).

Challenges here:

  1. This solution only works with certain PPRs that can be created with the clock dividers.
  2. Aligning Z to A will be impossible under certain conditionals as the minimum resolution of Z is PPR times worse than A.

To me, both suck, any suggestions?

  • 0
    TI__Mastermind 25956 points

    Hi Wil,

    Wil P. said:

    I am finding the biggest challenge to be properly aligning the index when the PPR of the QEP is large (for now, I am using 512, but I'd like a flexible solution).  Here are a few things that I have tinkered with, and some challenges that I have had:

    What are your constraints for the PPRs? Min/Max? Have you tried syncing the PWM modules? How many encoders are you emulating? If you're only trying to emulate 1 encoder, then using 2 ePWM modules would seem to work since you have ePWM1A/B, and ePWM2A. ePWM1A could be used as your index strobe, and ePWM1B could be used as your eQEPA signal. ePWM2A could be used as your QEPB signal. 

    The reason behind using ePWM1A as index and ePWM1B as QEPA, is that they should be properly aligned since they are within the same ePWM module and using the same CLK. 

    Best,

    Ryan Ma

  • 0 in reply to Ryan Ma
    Expert 1660 points

    My PPR is 512, I would like to up to 2048. I'm emulating one encoder but I would like the signals to be differential pairs to match the hardware inputs.

    All of the PWM modules are sync'ed, that is not the issue. The issue is reasonably "gating" the index pulse with the other pulses (see Figure 17-3 in SPRUHM8).

  • 0 in reply to Wil P.
    TI__Mastermind 25956 points

    Hi Wil,

    I am working on this with another subject matter expert to see if we can provide an alternative suggestion than the ones you have provided. I will provide a response by end of day tomorrow.

    Best,

    Ryan Ma

  • 0 in reply to Ryan Ma
    TI__Genius 17771 points

    Hi Will, 

    Can you clarify what you mean when you say that you want the signals to be differential pairs? The quadrature signals are typically single-ended. And since they're digital and not fast-switching, they're typically not particularly susceptible to common mode noise. 

    Do you instead mean that you want to ensure they are 90 deg offset?

    Regards,

    Peter

  • 0 in reply to Peter Luong1
    Expert 1660 points

    Peter,

    In my experience, its not uncommon for motor shaft QEP sensors to send their signals in differential pairs. Here is an example of one, if interested.

    With that said, I do not believe that differential is the issue. As I stated in the OP, I have A and B working fine (diff or not), that part is straight-forward. Gating Z is the part that I believe is tricky.

  • 0 in reply to Wil P.
    TI__Mastermind 25956 points

    Hi Wil,

    Are you using or planning to use any of the CLBs on the device? 

  • +1 in reply to Ryan Ma
    TI__Genius 17771 points

    Hi Will,

    Wil P. said:
    In my experience, its not uncommon for motor shaft QEP sensors to send their signals in differential pairs. Here is an example of one, if interested.

    Thanks for sending that link. Our EQEPs only support single ended QEP signals, so this is an interesting read and seems especially important in applications where the cable length will be quite long.

    Moving over to your inquiry, I discussed this with Ryan and an implementation we have thought of would use the configurable logic block (CLB) IP on the C2000 device. This CLB has connections to the EPWM so one solution we thought of would be to pass the QEPA_P signal through the CLB and have the CLB count the number of incoming signals. The CLB has an internal counter which you can configure to count to a specific number and then output a pulse of a defined length onto a GPIO pin once that count value has been received. Inputs into the CLB can be filtered for rising edge and falling edge detects so the CLB peripheral accurately counts the number of pulses received from the EPWM module.

    Regards,

    Peter

  • 0 in reply to Ryan Ma
    Expert 1660 points

    They are currently unused, and I am willing to try them out.

  • 0 in reply to Peter Luong1
    Expert 1660 points

    Peter,

    Sorry for being unclear, I have already handled the diff->single-ended conversion. This entire system is working in physical units with differential QEP inputs. This request is for creating an emulator of these large systems so I can better debug/development at my desk.

    That CLB solution seems really clever! For some reason, I thought that the CLB counters could only count up to 16.

    To summarize:

    1. Route the QEPA_P (or single-ended QEP_A) signal to a CLB input.
    2. Configure the CLB to use the edge-counting function (with filtering) to properly time the Z signal.
    3. Output QEPZ_P with a GPIO pulse of appropriate length.

    Instead of creating the Z signal with a GPIO pulse, would I be able use the CLB counter logic to trigger a One Shot on the PWM module that is configured for Z?  This way the differential part would be handled simply (EPWMLINK for TBPRD and CMP).

    Thanks again. Peter and Ryan!

  • 0 in reply to Wil P.
    TI__Mastermind 25956 points

    Hi Wil,

    The One shot on the PWM, are you meaning using the one shot trip zone configuration on PWM?  

    Best,

    Ryan Ma

  • 0 in reply to Wil P.
    TI__Genius 17771 points

    Hi Will,

    Wil P. said:
    That CLB solution seems really clever! For some reason, I thought that the CLB counters could only count up to 16.

    Yes we actually have three independent 32-bit counters within the CLB so even with the single counter, you would be able to support even the highest resolution encoders (I've yet to hear of a 1 billion count encoder)

    Wil P. said:
    Instead of creating the Z signal with a GPIO pulse, would I be able use the CLB counter logic to trigger a One Shot on the PWM module that is configured for Z? 

    I don't believe this is a feature on our PWMs, but you should be able to use another independent counter to control the on-time of this index pulse being generated by the CLB. There is a CPU-CLB FIFO interface which you can use to send data to the CLB counter so you can dynamically update this on-time as you update the frequency of the QEP signals

    Regards,

    Peter

  • 0 in reply to Ryan Ma
    Expert 1660 points

    Yes sir, I don't think I meant that though.  I will try what Peter suggested below.

    Thanks.

  • 0 in reply to Peter Luong1
    Expert 1660 points

    Very cool.  I am excited to try this.

    Thanks!

  • 0 in reply to Wil P.
    TI__Mastermind 25956 points

    Hi Wil,

    Sounds good, if any further questions come up feel free to post a new thread with your new question.

    I will go ahead and close this thread.

    Best,

    Ryan Ma