TIDM-DELFINO-ETHERCAT: Using Distributed Clock in EtherCAT network.

Santosh,

I did what you wrote. I generated SSC with patch provided in controlSuite for TMDSECATCNCD379D kit. I made some changes in defines to apply to launchxl-f28379d pins. And it is running, we were fighting with that in another thread and we did it successfully :)

But when I set DC mode in TwinCAT and then I am trying to send command like switch LED then I can't see any change in CCS debugger. Do I have to do something more to turn DC mode on?

BR,
Dawid.

Santosh,

I successfully ran DC mode - thanks to you :) You wrote that I should RELOAD DEVICES. I didn't do it before...

Now I am able to set cycle time to 500 us. Free run cycle is set to 1ms but in Advanced Settings I defined Sync 0 Cycle Time to 500 us.

My target is about 100 us but when I set time below 500 us, e.g. 400 us I can see "Synchronization error" in error list. Should I do something with shift times or Sync1?

I appreciate your help Santosh.

BR,
Dawid.

Dawid,

you don't have to necessarily do that, both SYNC0 and SYNC1 can be configured independent as needed, only thing you need to make sure is how the ISRs are handled in SW, while CPU is handling one ISR the second triggered one has to wait till first is serviced by default. Try disabling SYNC1 and see if it helps your experiments.

I'm not sure about the synchronization error you are talking about. Which PDI you are using ASYNC16 or SPI? There could be a limit to how low you can go with the SYNC time, with ASYNC16 the timing should be far better than SPI.

Hope this helps.

Best Regards

Santosh Athuru

Santosh,

sorry for late response, but today I am able to check this out in lab.
I am using ASYNC16. As you said, I disabled SYNC1 and I used only SYNC0. And still got the same. When SYNC0 is set with 500 us cycle it's still working but when I decrease it to e.g. 400 us I can see that error:

After that error RUN led on TIDM-DELFINO-ETHERCAT is blinking.

And it is my opinion too that ASYNC16 should has better timing limit that SPI.

What do you think about that?  Should I do something with shift time or there are another settings?

BR,
Dawid.

Santosh,

I did exactly what you wrote. I measured cycles within ISR and between them. Here are my screenshots:

Here is my SysPll config:

InitSysPll(XTAL_OSC,IMULT_40,FMULT_0,PLLCLK_BY_2);

I changed it from:

InitSysPll(/*INT_OSC2*/XTAL_OSC,IMULT_20,FMULT_0,PLLCLK_BY_1);

It seems that ISR is quite fast. What do you think about that?

BR,
Dawid.

Santosh,

I would like to provide you more information. Before you wrote how to check cycles in debug I was checking some time ranges by using timers. And I checked how often void APPL_Application(void); function is firening.
So:

1) in TwinCAT I set DC-mode with 600 us cycle time.
2) in main function I set timer 1 like that: CpuTimer1Regs.TPR.bit.TDDR = 199; so with 200 MHz clock 1 tick lasts 1 us.
3) in APPL_Application function I added that code at the beginning:

time = 0xFFFFFFFF - CpuTimer1Regs.TIM.all;
CpuTimer1Regs.TCR.bit.TSS = 1;     // 1 = Stop timer, 0 = Start/Restart Timer
CpuTimer1Regs.TCR.bit.TRB = 1;     // 1 = reload timer
CpuTimer1Regs.TCR.bit.TSS = 0;     // 1 = Stop timer, 0 = Start/Restart Timer

4) in Watch Expressions I added variable 'time'
5) load application and reload devices in TwinCAT.

As I said, cycle time is set to 600 us and while application running 'time' variable is about 600. Everything looks correct. Here is screen from watch expressions:

Next I checked cycles just like I checked cycles in ISR Sync0. I put breakpoints at the beginning and at the and of APPL_Application function. So, from beginning to the end APPL_Application function lasts about 77 cycles:

I checked time from returning of APPL_Application to next fireing this function and this is about 2720 cycles:

But at the same time, value of 'time' variable which was used to save measured time from timer decreased from 600 to 17! I have no idea why.

I thought it should be about 600 still. What do you think about this behaviour?

BR,
Dawid.

Santosh,

let's forget for a while about my previous questions. I would like ask you another very importan question.

I connected my Launchpad to another EtherCAT master based on AM335 sitara uP, it's TMDSICE3359 board from TI. In this master I increment processing value in each cycle, it's DataFromMaster register. In slave application on Launchpad in void APPL_Application(void) I wanted to save data in register, I mean sth like this:

void APPL_Application(void)
{
        register[i] = DatafromMaster0x7010.DatafromMaster;
        i++;
        //... another procedures
}

In watch expressions I can see that DatafromMaster register is incrementing fine, but in register array I can see that all register[i] values are equal to current Datafrommaster! I mean, it looks like APPL_Application functions fires much more faster than EtherCAT frame cycle. I thought APPL_Application fires each cycle time, e.g. 1ms or 500 us, generally depends on master.

Situation I described looks like this:

DatafromMaster = 0x01;

register[0] = 0x01, register[1] = 0x01, register[2] = 0x01, ... , register[n-1] = 0x01;

---sending frame, cycle time elapsed---

DatafromMaster = 0x02;

register[0] = 0x02, register[1] = 0x02, register[2] = 0x02, ... , register[n-1] = 0x02;

I thought it should be:

DatafromMaster = 0xFF;

register[0] = 0, register[1] = 0x01, ... , register[0xFE] = 0xFF.

What do you think about that?

BR,
Dawid.