IWR1443: large transfer overhead via SPI driver

Hi Andrey,

Would you be able to post a scope shot of this?

This is a good question and I will need to interface with my team before providing a response. I will try to have a response in the next few days.


Cheers,
Akash

Hi Andrey,

Do you have a scope shot you can provide or some kind of data that supports your measurement?


Cheers,
Akash

Sorry for the delay.

At first, our SPI init:

-------------------------------------------------------------------------------------------------------------------

    SPI_Params      params;
    SPI_Transaction spiTransaction;
    SPI_Params_init(&params);
    params.mode  = SPI_MASTER;
    params.pinMode=SPI_PINMODE_3PIN;
//    params.csHold=1U;
    params.frameFormat=SPI_POL0_PHA1; //SPI_POL0_PHA1
    params.dataSize=8U;
    params.u.masterParams.c2tDelay=2U;
    params.u.masterParams.t2cDelay=2U;
    params.u.masterParams.wDelay=0U;
    params.u.masterParams.bitRate = (uint32_t)(16*1000000);
    params.u.masterParams.numSlaves = 1U;
    params.dmaHandle=gMmwMCB.gDmaHandle;
    params.dmaEnable=0;
    params.u.masterParams.slaveProf[0].chipSelect = 0U;
    params.u.masterParams.slaveProf[0].ramBufLen = (uint8_t)(MIBSPI_RAM_MAX_ELEM);
    params.u.masterParams.slaveProf[0].dmaCfg.txDmaChanNum =1U;
    params.u.masterParams.slaveProf[0].dmaCfg.rxDmaChanNum =0U;
    params.u.masterParams.slaveProf[1].chipSelect = 1;
    params.u.masterParams.slaveProf[1].ramBufLen = (uint8_t)MIBSPI_RAM_MAX_ELEM/2;
    params.u.masterParams.slaveProf[1].dmaCfg.txDmaChanNum =3U;
    params.u.masterParams.slaveProf[1].dmaCfg.rxDmaChanNum =2U;
    gMmwMCB.spiHandle = SPI_open(0, &params);
    if (!gMmwMCB.spiHandle) {
        DebugAssert (0);
    }

//    SPI transaction settings
    GPIO_write (SOC_XWR14XX_GPIO_30, 1U);       // CS enable
    gMmwMCB.spiTransaction.count = 1; // default value
    gMmwMCB.spiTransaction.txBuf = NULL;
    gMmwMCB.spiTransaction.rxBuf = NULL;
    gMmwMCB.spiTransaction.slaveIndex = 0;  // external RAM spi channel

Sending...
----------------------------------------------------------------------------------
trans->count = 10000;
trans->rxBuf=NULL;
trans->txBuf = p;

errCode |= SPI_transfer(gMmwMCB.spiHandle, trans);

At end, scope shot

Upper ray - CS

Lower ray - CLK (highly discontinuous !)

Hi Andrey,

Can you confirm you are using the IWR1443BOOST or are you using the IWR1443 on a custom board?


Cheers,
Akash

Hello.

It's a custom board with IWR1443 and 23K256(on SPI) .

On other board, with another SoC, same RAM works via SPI without such delays well. As slave, why not...

Hi Andrey,

I am currently discussing your issue with our hardware team, we will try to have an answer for you early next week.


Cheers,
Akash

Hi,

Looking at the scope shot, the time between the data transfer blocks seems to vary. Is there some delay on the slave side?
Also are the individual blocks of data shown on the scope sending with the expected timing?

Regards,
Charles O

Hello.

I guess, choosen SPI protocol don't to let the slave paused a master.

Besides our slave chip (23K256 SRAM) can't do that.

And on other master SoC in another working scheme - it runs continously (10K data block).

Of course, if all delays subtract from measurement time then will be about right (expected) period.

Thanks.