Problem with timing LPSPI (i.MXRT1062)

[arnipol]

I have a problem with LPSPI (i.MXRT1062). It's about starting the transmission (Master).

Starting PSC + CLOCK begins after ~ 2 us after writing data  to registry (LPSPI->TDR = data or TCR.TXMSK = 1). Is this normal behavior?

On the picture: TDR is updated with the falling edge of the yellow signal. Clock starts 2u later.

 

[kerryzhou]

Hi arni pol,

   Thanks for your updated information.

   In fact, the test result wave delay time is not very correct, as you know, the gpio toggle also needs time.

   About DbgConsole_TryGetchar influence your interrupt, do you try to disable it, any improvement?

Kerry

-------------------------------------------------------------------------------
Note:
- If this post answers your question, please click the "Mark Correct" button. Thank you!

- We are following threads for 7 weeks after the last post, later replies are ignored
Please open a new thread and refer to the closed one, if you have a related question at a later point in time.
-------------------------------------------------------------------------------

[arnipol]

Time Tlead (3)  I can decrease to few nanoseconds.

The problem is to decrease time between command putting data to SPI and  SPI_PCS.

See below on picture1 - the red 1.8us. I don't know why SPI starting sendig data so late.

Picture2 -  blue line - PCS

                                      LPSPI_WriteData(EXAMPLE_LPSPI_MASTER_BASEADDR, masterTxData[masterTxCount]);

                  yellow line ->  GPIO_PinWrite(BOARD_INITPINS_CONVST1_GPIO, BOARD_INITPINS_CONVST1_PIN, 1); // yeallow signal

Picture1

Picture2

[kerryzhou]

Hi arni pol,

   I can understanding you now.

   Please test the following situation, whether you can decrease the time or not:

1.  LPSPI_WriteData(EXAMPLE_LPSPI_MASTER_BASEADDR, masterTxData[masterTxCount]);

  This code use the register :

LPSPI1->TDR = masterTxData[masterTxCount]);

2. You can set more high module clock:

Any change or not?

This time is determined by the chip and the related core clock and module clock.

Have a great day,
Kerry

-------------------------------------------------------------------------------
Note:
- If this post answers your question, please click the "Mark Correct" button. Thank you!

- We are following threads for 7 weeks after the last post, later replies are ignored
Please open a new thread and refer to the closed one, if you have a related question at a later point in time.
-------------------------------------------------------------------------------

[arnipol]

The clocks are set correctly, but I finally found a problem. Because I want only read the data from SPI,  i iniciate transmission  by setting the TCR.TXMSK bit = 1. Data is not loaded into the FIFO and the clock starts almost immediately (~ 300ns). Putting data to LPSPI-> TDR causes a much longer delay.

The second problem was delay of port interruption in wich I started the spi clock.

 Although the interruption had the highest priority, the delay of two microseconds still occurred.
The problem was cased by debug console, function DbgConsole_TryGetchar(in main loop)    which blocks all interruptions  by ~2us. Is' it necesery to blocked all interruption? I understand that it protects it's own  resources but sometimes causes the other problems (like in my case).

Anyway, thank You for your help.

Best regards,

Arkadiusz Polak

[arnipol]

The yellow signal comes from the ADC converter and triggers the interruption. Inside the interruption, the lpspi clock is initialized to receive data for the next int.

I have checked nxp sample. A little changes i've made are: increase clock to 20Mhz  and insert the <pin triger>

....

/*Fill up the TX data in FIFO */
while ((LPSPI_GetTxFifoCount(EXAMPLE_LPSPI_MASTER_BASEADDR) < g_masterFifoSize) &&
(masterTxCount - masterRxCount < g_masterFifoSize))
{

/*Write the word to TX register*/

LPSPI_WriteData(EXAMPLE_LPSPI_MASTER_BASEADDR, masterTxData[masterTxCount]);

GPIO_PinWrite(BOARD_INITPINS_CONVST1_GPIO, BOARD_INITPINS_CONVST1_PIN, 1); // yeallow signal

++masterTxCount;

.....

Time between between LPSPI_WriteData and LPSPI_CS is about 1.8 us. Is't possible to decrease it?

Clock start's more late,  but this is due to wrong calculatin of delay (in SDK 2.7) 

[kerryzhou]

Hi arni pol,

  Could you please also add the SPI_CS pin? Do you have more channels in your oscilloscope?

  Do you want to decrease the time between the SPI_CS and the CLK ?

  

   Do you want to decrease the item 3 time?

Have a great day,
Kerry

-------------------------------------------------------------------------------
Note:
- If this post answers your question, please click the "Mark Correct" button. Thank you!

- We are following threads for 7 weeks after the last post, later replies are ignored
Please open a new thread and refer to the closed one, if you have a related question at a later point in time.
-------------------------------------------------------------------------------

[kerryzhou]

Hi arni pol,

  Please tell me yellow signal is what in your picture ? LPSPI_CS?

 Do you test the SDK code, eg:

SDK_2.7.0_EVK-MIMXRT1060\boards\evkmimxrt1060\driver_examples\lpspi

 The same wave?

  Please tell me your wave signal in your picture at first, then test the SDK code SPI wave, then I can help you to check more details.

Wish it helps you!

If you still have questions about it, please kindly let me know.

Have a great day,
Kerry

-------------------------------------------------------------------------------
Note:
- If this post answers your question, please click the "Mark Correct" button. Thank you!

- We are following threads for 7 weeks after the last post, later replies are ignored
Please open a new thread and refer to the closed one, if you have a related question at a later point in time.
-------------------------------------------------------------------------------