TMDSCNCD28379D: Realtime PWM with multiple controller

Hi David!

Could you help expand a bit on what you are trying to do? What is the reasoning for using SCI here?

Do you mean that you are trying to send data from one C2000 to another and have the other device(s) update the EPWM signals based upon what is being sent via SCI?

Best Regards,

Allison

Hi Allison!

I want to send data containing calculated PWM signals to a C2000, which sends the data to another C2000 and so on. The microcontrollers should output the PWM from the data.

It is important that the PWM signals output by the controllers are synchronised with each other.

Best Regards,

David

Hi David, 

Do you mean that you want to use the SCI data lines as PWM signals? Or need to communicate information to be used by the PWMs? Just wanting to be sure I'm interpreting the wording here correctly!

Does this past post describe a similar situation to yours at all? https://e2e.ti.com/support/microcontrollers/c2000-microcontrollers-group/c2000/f/c2000-microcontrollers-forum/844385/tms320f28379d-use-output-of-epwm-and-synchronize-them 

You'll need to leverage the external sync source (EXTSYNCIN) options to sync PWMs across devices. 

Best Regards,

Allison

Hello Allision,

thanks for your message. Yes, as David already said, we want to communicate information to be used by the PWMs. Moreover, due to the high voltage isolation between each controller, the communication is done via fiber optical lines. Therefore, the synchronization should be done also with the communication lines and not with an extra fiber line on the EXTSYNCIN signal.

We are also planning to use the F28388 and have the option of using Ethercat.

Do you have a solution to synchronize with the communication line?

Best regards,

Patrick

Hi Patrick,

1. Do you need to sync multiple times? Or just once at the beginning of communication?
2. What communication protocol are you planning on using? 
3. Are all the C2000 peripheral devices receiving the controller device's transmission at the same time? Or is there some nondeterministic delay between when each device receives the communication signal?

One idea is to leverage the XBARs and CLB. You can have INPUT XBAR read the communications line reading idle by default and se this as the initial CLB state machine (e.g. state 0). As soon as the idle communications line becomes active, the CLB can detect this and go to another state (e.g. state 1) where the EPWM is starts up. If the communication transmission is sent to all the devices at the same time, the EPWMs would be synced. 

You could also look into changing the GPIO settings of the fiber optic GPIOs to use OUTPUT XBAR (EXTSYNCOUT).  

Best Regards,

Allison

Hello Allison,

thanks Allison for your quick reply.

Normally, one sync at the beginning should be fine but I think that due to the long operating time a synchronization after a predefined time is necessary.

We would like to use a daisy chain to reduce the number of wires. A small delay between each module is acceptable. We are not committed to a specific communication protocol. We plan to use something similar to UART.

Thanks for your solution. We will evaluate it soon.

Best Regards,

Patrick

Hi Patrick,

Syncing once in the beginning should be fine if you are starting them at the same time by communicating to all devices at the same time. Issuing a sync later would likely need to be implemented in another way though - could be in hardware if you can spare the extra connection. Otherwise through software. 

By daisy chaining are you implying you send it through one device at a time? Or would the communication reach each device essentially simultaneously?

And yes let me know if you have thoughts on the proposal. 

Best Regards,

Allison

Hello Allison,

Ideally, each device will communicate simultaneously, but due to hardware, there is a delay between each controller that should be corrected by the controller.

Best regards,

Patrick