This is my first time using QSPI which was partly the cause of my confusion.  I couldn't figure out if it was something with my custom hardware or my firmware that was causing me so much trouble.  I'm using a STM32WB55REV for my two custom boards, one is the sender, and one is the receiver, the receiver is the one connected to a Riverdi display with the BT817Q.

I do not see in your code to which address the command is written to.

numberCommand.Instruction = 0xB0; // Get the first byte, should 0xB0 (0x10) to write

And the rest?

If I understand correctly, the QSPI is broken into Command and Address (excuse the variable names, I've been copy/pasting and have to come back to refactor everything, just trying to get something working initially):

numberCommand.Instruction = 0xB0; // Get the first byte, should 0xB0 (0x10) to write
numberCommand.Address     = 0x2578;

Those two lines are equivalent to these lines in your library:

spi_transmit((uint8_t) 0xB0U); /* high-byte of REG_CMDB_WRITE + MEM_WRITE */
spi_transmit((uint8_t) 0x25U); /* middle-byte of REG_CMDB_WRITE */
spi_transmit((uint8_t) 0x78U); /* low-byte of REG_CMDB_WRITE */

Then it sets up the EVE command and the value:

uint8_t cmdArray[8] = {
(uint8_t)(CMD_SETROTATE & 0x000000ff),
(uint8_t)(CMD_SETROTATE >> 8),
(uint8_t)(CMD_SETROTATE >> 16),
(uint8_t)(CMD_SETROTATE >> 24),
/* Rotation */
(uint8_t)(rotation & 0x000000ff),
(uint8_t)(rotation >> 8),
(uint8_t)(rotation >> 16),
(uint8_t)(rotation >> 24)
};


CS likely is hardware controlled?

Correct, the CS is completely out of my hands which is making it difficult to create the same kind of flow of your library where you setup the _burst commands because I haven't figured out how to keep the CS line pulled low while transmitting the rest of the EVE commands. 

What does this look like on the logic analyzer? The whole sequence for this function and perhaps some bits and pieces before and after.

I've captured waveforms with my Digilent AD 3 and Waveforms software, but they don't really have a QSPI decoder, so I had no idea what I was looking at.  As per the suggestion from BRT, I connected an oscilloscope to the QSPI lines and they didn't look well-formed which is what prompted me to change the skew rate, again suggested by BRT.  This has seemed to resolve the issue for me.

Kindest regards.

Hello,

Thank you for your response.  After more research, I think the issue is in the set rotate command, nothing to do with reading from the flash.  I went through and put this line of code after pretty much everything, so I could see how much space was available:

uint16_t space;
space = TFT_qspi_read16(REG_CMDB_SPACE);

The rotate command always returns 4084 when the display doesn't work because it gets into the infinite loop, checking if REG_CMD_READ == REG_CMD_WRITE or if I use the TFT_Busy, it always returns EVE_FIFO_HALF_EMPTY.

Is there a reason why rotating the screen would have the REG_CMD_SPACE equal to 4084?  Every other call to TFT_qspi_read16(REG_CMDB_SPACE) always returns 4092. 

space = TFT_qspi_read16(REG_CMDB_SPACE); //Always 4092
TFT_qspi_cmd_rotate(1);
space = TFT_qspi_read16(REG_CMDB_SPACE); //When it fails, always 4084

void TFT_qspi_cmd_rotate(uint32_t rotation)
{
QSPI_CommandTypeDef numberCommand = {0};

// Set up the command structure
numberCommand.InstructionMode   = QSPI_INSTRUCTION_4_LINES;    // Instruction sent over 1 line
numberCommand.Instruction       = 0xB0; //Get the first byte, should 0xB0 (0x10) to write
numberCommand.AddressMode       = QSPI_ADDRESS_4_LINES;        // Address sent over 1 line
numberCommand.AddressSize       = QSPI_ADDRESS_16_BITS;        // Try using 16-bit address
numberCommand.Address           = 0x2578;            // Mask off the first byte and keep the last two bytes
numberCommand.DataMode          = QSPI_DATA_4_LINES;            // Data mode for receiving ID
numberCommand.NbData    = 8;                          // Expecting to receive 4 bytes
numberCommand.DummyCycles       = 0;                          // No dummy cycles
numberCommand.DdrMode           = QSPI_DDR_MODE_DISABLE;      // Disable DDR mode
numberCommand.SIOOMode          = QSPI_SIOO_INST_EVERY_CMD;    // Send instruction every command

uint8_t cmdArray[8] = {
(uint8_t)(CMD_SETROTATE & 0x000000ff),
(uint8_t)(CMD_SETROTATE >> 8),
(uint8_t)(CMD_SETROTATE >> 16),
(uint8_t)(CMD_SETROTATE >> 24),
/* Rotation */
(uint8_t)(rotation & 0x000000ff),
(uint8_t)(rotation >> 8),
(uint8_t)(rotation >> 16),
(uint8_t)(rotation >> 24)
};

if (HAL_QSPI_Command(&hqspi, &numberCommand, HAL_QSPI_TIMEOUT_DEFAULT_VALUE) != HAL_OK)
Error_Handler();

if (HAL_QSPI_Transmit(&hqspi, cmdArray, HAL_QSPI_TIMEOUT_DEFAULT_VALUE) != HAL_OK)
Error_Handler();

return;
}

Any suggestions would be greatly appreciated.

Kindest regards.

Hello,

Thanks for your response.

I haven't tried using SPI thus far, I'll have to dig out a board and give it a try.  While trying to debug this situation the only thing I've found so far is that sometimes the flash initialization fails:

if(TFT_init_flash() != E_OK)
Error_Handler();

To check for possible co-processor errors, I added this code placing a breakpoint at the bottom:

if(TFT_qspi_read16(REG_CMD_READ) == 0xFFF){
uint8_t Offset = 0;
uint8_t ErrChar;
uint8_t result[128 + 1]; // Buffer to hold the full string (including null terminator)
int result_index = 0;
do
{
// Get the error character and display it
ErrChar = TFT_qspi_read8(EVE_RAM_ERR_REPORT + Offset);
// Only append valid characters (non-null)
if (ErrChar != 0 && result_index < 128) {
result[result_index++] = ErrChar;
}

Offset++;
}while ( (ErrChar != 0) && (Offset < 128) ); // when the last stuffed character was null, we are done

// Null-terminate the string
result[result_index] = '\0';

// Now the full string is in 'result', which you can use or print
printf("Captured string: %s\n", result); //Line with breakpoint
}

It never hits the breakpoint, even when the screen just has the backlight, no graphics at all.

Thanks for the link to the latest programmer's guide, I was using an older version and didn't initially see the EVE_RAM_ERR_REPORT, this has since come in quite handy while trying to get custom fonts working via QSPI.

If I can do anything to help facilitate a solution, please let me know.

Kindest regards.

WOW, I can't believe it, I actually got it to work, very cool!  I set the prescaler to be 80, which brings the clock up to roughly 200khz.  My code looks like this now:

HAL_GPIO_WritePin(QSPI1_RST_GPIO_Port, QSPI1_RST_Pin, GPIO_PIN_RESET);
HAL_Delay(21); /* minimum time for power-down is 5ms */
HAL_GPIO_WritePin(QSPI1_RST_GPIO_Port, QSPI1_RST_Pin, GPIO_PIN_SET);
HAL_Delay(21); /* minimum time to allow from rising PD_N to first access is 20ms */

//External Crystal
TFT_sendCmd(CMD_CLKEXT, 0x00);
HAL_Delay(100);
/* set clock to 72 MHz */
TFT_sendCmd(CMD_CLKSEL, 0x46);
/* start EVE */
TFT_sendCmd(CMD_ACTIVE, 0x00);
/* give EVE time to power up */
HAL_Delay(40);

/* read the ID */
uint8_t regId = TFT_readId(REG_ID);
while(regId != 0x7c)
    regId = TFT_readId(REG_ID);

/* ensure CPUreset register reads 0 signaling it's ready */
uint8_t cpuState = TFT_read8(REG_CPURESET);
while ( cpuState != 0x00)
    cpuState = TFT_read8(REG_CPURESET);

/* switch to QSPI */
TFT_write8(REG_SPI_WIDTH, 0x02);
if(TFT_qspi_read8(REG_SPI_WIDTH) != 0x02)
    Error_Handler();

/* tell EVE that we changed the frequency to 72MHz */
TFT_qspi_write32(REG_FREQUENCY, 72000000);
/* read the frequency to verify */
if(TFT_qspi_read32(REG_FREQUENCY) != 72000000)
    Error_Handler();

/* turn off back light */
TFT_qspi_write8(REG_PWM_DUTY, 0);
/* active display width */
TFT_qspi_write16(REG_HSIZE, 800);
/* verify the value was set */
if(TFT_qspi_read16(REG_HSIZE) != 800)
    Error_Handler();

I've attached my capture.  I'm currently using WaveForms which I don't care for much but it's what I have right now.  The capture is pretty close to what you posted for the REG_HSIZE.

I have a lot of code yet to write to get something to show up on the display, but this is a HUGE start.

Thanks so much for your assistance.

Kindest regards

This is GREAT information, thank you! 

My SCK is VERY slow at this point, just trying to keep variables to a minimum while I try to get it working.  My QSPI clock is, 16Mhz and the prescaler is set at 250.

I've actually been using SPI traces to get my QSPI commands sorted so if you could share a working QSPI trace I could reference to try and match, I think that would be very helpful, thank you!

I will implement your suggestions and if I'm still having issues I will get a trace for analysis.

Again, thank you for your detailed explanation.

Kindest regards.

Hello,

I currently have Rudolphs library and the Riverdi_EVE library running via SPI.  All that is working just fine.  I'm using custom hardware with an STM32WB55REV6 on it and the QUADSPI bus is what's connected to the Riverdi display.  I tried using the QUADSPI in single mode but have failed miserably so thought I'd try using QUAD SPI instead.

Looking at the datasheet, 4.1 Quad SPI Host Interface, I believe I need to write 0x04 to the REG_SPI_WIDTH.  At which point in the initialization process, documented in the BRT_AN_033_BT81X-Series-Programming-Guide, section 2.4 Initialization Sequence during Boot Up?  Should REG_SPI_WIDTH be set after setting everything up or is it the first thing that should be set after the RST/PD sequence?

Any other things that might be helpful getting it up and running via QUAD SPI?

Any insight would be greatly appreciated.

Could I clarify which version of EVE Asset Builder you care currently utilising?

I'm using EVE Asset Builder v2.10.2

Just as a note, if a erase does succeed the blob would need re-programmed into the flash IC before it can be detected. It may be that an erase has completed and the subsequent write has failed leading to he blob note being present.

EVE Asset Builder is reporting that it successfully erased the data but it's a false positive.  I can plug in the FT232HQ and it works fine identifying the blob and calculating the speed.

Anything else you need me to try please let me know.

Kindest regards

Hello, I was wondering if you could share your wiring connections.  I'm struggling to get the UMFT4222EV to connect to a Riverdi BT817Q.  My connections are as follows:

UMFT4222EV     ->     BT817Q
JP5.9                ->     SCLK
JP5.8                ->     MISO
JP5.7                ->     MOSI
JP5.5                ->     IO.2
JP5.4                ->     IO.3

JP4.11              ->     -----     10k Resistor -> +3V3
JP4.9                ->     CS        4.7k Resistor -> +3V3
JP4.6                ->     INT
JP4.4                ->     RST/PD

When in EAB and I try to Detect, I mostly get the following:

Detecting flash...
Found device: FT4222 A (A)
Host: FT4222

Flash size: 64 MBytes

Blob is NOT present.

Flash can switch to full speed.
END

Detection fail!

Occasionally, I get this:

Detecting flash...
Found device: FT4222 A (A)
Host: FT4222

Flash size: 64 MBytes

Blob is NOT present.

Flash can switch to full speed.
END

BEGIN
Measure reading speed...

Reading speed: 17.88 MB/s
Reading 10 MBytes in 0.56 seconds
END

In either case, it never finds the blob.  I know the blob is present because I can program it perfectly fine with an FT232H.

Is something connected incorrectly?  I've been trying to use the ME817EV schematic to try and match things up, which I think I have (using the pin mapping above).

Any insight would be greatly appreciated.

Kindest regards.