Quad-SPI DDR indirect read mode dummy cycles count

[axlrose21]

Hello ST community,

I have some troubles with dummy cycles using Quad-SPI interface in DDR mode on STM32H743 and STM32F746.

I am trying to read data from IS25LP080D QSPI flash with FAST READ QUAD IO DTR MODE OPERATION (0xED command). It is 1-4-4 mode.
I am having trouble with number of dummy cycles. From IS25LP080D datasheet there must be 6 dummy sclk cycles between address and data. But to get 6 dummy cycles I must set cmd.DummyCycles = 3; in my program. Why there are 3 extra dummy cycles on diagram?

My read function for HAL is below. In this function I set cmd.DummyCycles = 3.

IS25LP_Status_t IS25LP_ReadQuadDDR(QSPI_HandleTypeDef *hqspi, uint8_t *buf, uint32_t addr, uint32_t size)
{
    QSPI_CommandTypeDef cmd = {0};

    cmd.InstructionMode    = QSPI_INSTRUCTION_1_LINE;
    cmd.Instruction        = IS25LP_FAST_READ_QUAD_IO_DDR;
    cmd.AddressMode        = QSPI_ADDRESS_4_LINES;
    cmd.AddressSize        = QSPI_ADDRESS_24_BITS;
    cmd.Address            = addr;
    cmd.AlternateByteMode  = QSPI_ALTERNATE_BYTES_NONE;
    cmd.AlternateBytesSize = QSPI_ALTERNATE_BYTES_24_BITS;
    cmd.AlternateBytes	   = 0x0;
    cmd.DummyCycles        = 3;
    cmd.DataMode           = QSPI_DATA_4_LINES;
    cmd.NbData             = size;
    cmd.DdrMode            = QSPI_DDR_MODE_ENABLE;
    cmd.DdrHoldHalfCycle   = QSPI_DDR_HHC_HALF_CLK_DELAY;
    cmd.SIOOMode           = QSPI_SIOO_INST_EVERY_CMD;

    if (HAL_QSPI_Command(hqspi, &cmd, QSPI_TIMEOUT) != HAL_OK) {
        return IS25LP_ERROR;
    }

    if (HAL_QSPI_Receive(hqspi, buf, QSPI_TIMEOUT) != HAL_OK) {
        return IS25LP_ERROR;
    }

    return IS25LP_OK;
}

In my main.c I use this function to read 8 bytes of data:

status = IS25LP_ReadQuadDDR(&hqspi, readBuf, 0x1000, 8);

And I am getting correct flash data:
First 8 bytes (DDR): 00 01 02 03 04 05 06 07

DCYC field in CCR register is 0x3, so there must be only 3 dummy cycles.

But on time diagram I can see 6 dummy cycles:

So I got 3 extra dummy cycles on STM32H743 board.

My question is: why there are extra dummy cycles? How should I calculate there number?

I also tried this code on STM32F746 board and I got 1 extra dummy cycles, but not 3.

Thank you for any thoughts!

I'm attaching my main.c and is25lp_mem.c for the reference.

Processor: STM32H743ZI

Board: Nucleo-H743ZI

Driver version: STM32Cube MCU Package for STM32H7 1.12.0

Compiler: GNU Tools for STM32 (12.3.rel1)

Cube MX 6.13.0