STM32F767 SRAM problem

megahercas6 · ‎2017-07-28

Posted on July 28, 2017 at 09:25

Hello, i made my own devkits with huge altera and STM32F4/7 micro-controllers.

As usual, STM32F479 performing very well, it uses 26b ADR, 32b data lines directly to FPGA.

But for some reason code generated by STM32CubeMX does not want to work. I mapped 4 banks of SRAM 16Mx32b to SRAM controllers, but they don't work, at all.

I am using this code to test it:

&sharpdefine Bank1_SRAM1_ADDR ((uint32_t)0x60000000) // FPGA

volatile uint32_t FMC_ADR_R = 0;

volatile uint32_t FMC_ADR_W = 0;

volatile uint32_t FMC_R = 0;

volatile uint32_t FMC_W = 0;

while (1)

{

*(uint32_t *) (Bank1_SRAM1_ADDR+FMC_ADR_W*4) = FMC_W;

*(uint32_t *) (Bank1_SRAM1_ADDR+0) = 0;

FMC_R = *(uint32_t *) (Bank1_SRAM1_ADDR+FMC_ADR_R*4);

}

By using debugger, i am changing FMC_ADR_R, FMC_ADR_W, FMC_W values, and CS goes only single time low during init, after that it keeps high. Adding more delay did not helped between read writes, so i don't know whats hoing on.

Also, in migration documentations i read that it only have single sram controller ( chip select) while STM32F4XX does have 4, whats up to that ? so why CubeMX is very happy generating code for all 4 chip selects ?

http://www.st.com/content/ccc/resource/technical/document/application_note/73/76/21/47/ef/d0/4c/16/DM00164538.pdf/files/DM00164538.pdf/jcr:content/translations/en.DM00164538.pdf

Page 24

Any one knows whats going on ?

My sram init looks like this:

FMC_NORSRAM_TimingTypeDef Timing;

/** Perform the SRAM1 memory initialization sequence

*/

hsram1.Instance = FMC_NORSRAM_DEVICE;

hsram1.Extended = FMC_NORSRAM_EXTENDED_DEVICE;

/* hsram1.Init */

hsram1.Init.NSBank = FMC_NORSRAM_BANK1;

hsram1.Init.DataAddressMux = FMC_DATA_ADDRESS_MUX_DISABLE;

hsram1.Init.MemoryType = FMC_MEMORY_TYPE_SRAM;

hsram1.Init.MemoryDataWidth = FMC_NORSRAM_MEM_BUS_WIDTH_32;

hsram1.Init.BurstAccessMode = FMC_BURST_ACCESS_MODE_DISABLE;

hsram1.Init.WaitSignalPolarity = FMC_WAIT_SIGNAL_POLARITY_LOW;

hsram1.Init.WaitSignalActive = FMC_WAIT_TIMING_BEFORE_WS;

hsram1.Init.WriteOperation = FMC_WRITE_OPERATION_DISABLE;

hsram1.Init.WaitSignal = FMC_WAIT_SIGNAL_DISABLE;

hsram1.Init.ExtendedMode = FMC_EXTENDED_MODE_DISABLE;

hsram1.Init.AsynchronousWait = FMC_ASYNCHRONOUS_WAIT_DISABLE;

hsram1.Init.WriteBurst = FMC_WRITE_BURST_DISABLE;

hsram1.Init.ContinuousClock = FMC_CONTINUOUS_CLOCK_SYNC_ASYNC;

hsram1.Init.WriteFifo = FMC_WRITE_FIFO_DISABLE;

hsram1.Init.PageSize = FMC_PAGE_SIZE_NONE;

/* Timing */

Timing.AddressSetupTime = 5;

Timing.AddressHoldTime = 3;

Timing.DataSetupTime = 5;

Timing.BusTurnAroundDuration = 0;

Timing.CLKDivision = 0;

Timing.DataLatency = 0;

Timing.AccessMode = FMC_ACCESS_MODE_A;

/* ExtTiming */

if (HAL_SRAM_Init(&hsram1, &Timing, NULL) != HAL_OK)

{

_Error_Handler(__FILE__, __LINE__);

}

Also, i was messing with timing, it does not help at all :\

#error #fmc #sram #problem #cubemx

Vangelis Fortounas · ‎2017-07-28

Posted on July 28, 2017 at 17:56

Hello!

I dont have your hardware to reproduce the error but in your initialisation you put FMC_WRITE_OPERATION_DISABLE

This is 'write protect ' functionality.

*(uint32_t *) (Bank1_SRAM1_ADDR+FMC_ADR_W*4) = FMC_W;

*(uint32_t *) (Bank1_SRAM1_ADDR+0) = 0;

You try to Write.

It will help the community if you post the .IOC file

megahercas6 · ‎2017-07-28

Posted on July 28, 2017 at 18:02

Well spotted, while in cubemx it was disabled.

Also i did use function to disable it ( called from sram driver, don't remember prototype), but had no effect, but all right, will try to edit my sram init file, maybe it will help.

But it should be reading, but it is not, stays high all the time, while with STM32F4 i get around 20MHz

Vangelis Fortounas · ‎2017-07-28

Posted on July 28, 2017 at 18:15

check your call Timing initialisation function

megahercas6 · ‎2017-07-31

Posted on July 31, 2017 at 08:01

Just like i suspected, after generating sram init code i change write to enable, so this is not a problem.

FMC_NORSRAM_TimingTypeDef Timing;

/** Perform the SRAM1 memory initialization sequence

*/

hsram1.Instance = FMC_NORSRAM_DEVICE;

hsram1.Extended = FMC_NORSRAM_EXTENDED_DEVICE;

/* hsram1.Init */

hsram1.Init.NSBank = FMC_NORSRAM_BANK1;

hsram1.Init.DataAddressMux = FMC_DATA_ADDRESS_MUX_DISABLE;

hsram1.Init.MemoryType = FMC_MEMORY_TYPE_SRAM;

hsram1.Init.MemoryDataWidth = FMC_NORSRAM_MEM_BUS_WIDTH_32;

hsram1.Init.BurstAccessMode = FMC_BURST_ACCESS_MODE_DISABLE;

hsram1.Init.WaitSignalPolarity = FMC_WAIT_SIGNAL_POLARITY_LOW;

hsram1.Init.WaitSignalActive = FMC_WAIT_TIMING_BEFORE_WS;

hsram1.Init.WriteOperation = FMC_WRITE_OPERATION_ENABLE;

hsram1.Init.WaitSignal = FMC_WAIT_SIGNAL_DISABLE;

hsram1.Init.ExtendedMode = FMC_EXTENDED_MODE_DISABLE;

hsram1.Init.AsynchronousWait = FMC_ASYNCHRONOUS_WAIT_DISABLE;

hsram1.Init.WriteBurst = FMC_WRITE_BURST_DISABLE;

hsram1.Init.ContinuousClock = FMC_CONTINUOUS_CLOCK_SYNC_ONLY;

hsram1.Init.WriteFifo = FMC_WRITE_FIFO_DISABLE;

hsram1.Init.PageSize = FMC_PAGE_SIZE_NONE;

/* Timing */

Timing.AddressSetupTime = 10;

Timing.AddressHoldTime = 10;

Timing.DataSetupTime = 10;

Timing.BusTurnAroundDuration = 1;

Timing.CLKDivision = 1;

Timing.DataLatency = 1;

Timing.AccessMode = FMC_ACCESS_MODE_A;

/* ExtTiming */

if (HAL_SRAM_Init(&hsram1, &Timing, NULL) != HAL_OK)

{

_Error_Handler(__FILE__, __LINE__);

}

But why is it ? Even original example with sram does not generate any signals, maybe it is because IAR ARM 8.2 ?

I did try to change optimization levels, no response, same chip select behavior. Also i did check gpio init code, it does enable chip select pin, and it does enable that pin clock ( even if i have only chip select enabled as AF, it will generate signals, while complete SRAm interface will be dead, i understand that)

Also, it does read single time correctly, in FPGA i have current FW version number, and it does read it, but only single time. Strange thing is that address should be 0, but it reading that number at address 1 (((uint32_t)0x60000004)

But again, why only single time ? :\

Attached my complete generated code from STM32CubeMx, maybe some one could take a look, or it is another fabulues times of STM32CubeMx not working ?

For the love of god, ST, produce SPL libraries for F7 and H7 micro controllers, since i am afraid of using them, and will stick to STM32F4 series instead

:(

And they are getting quite slow in comparison of H7 and F7

________________

Attachments :

STM32F767NIH.rar : https://st--c.eu10.content.force.com/sfc/dist/version/download/?oid=00Db0000000YtG6&ids=0680X000006HyXK&d=%2Fa%2F0X0000000b9D%2FGPAp9lsvF_kfDJZlfHVi5c8Gvv5DvyxtJ_XS0E1ajYw&asPdf=false

megahercas6 · ‎2017-08-02

Posted on August 02, 2017 at 15:42

The correct answer i was enabled D and I cache at the same time with accelerator, and it cause this.

Please CubeMX programmers, produce fail safe CPU parameters, so if one is selected, others could not be enabled !!

T J · ‎2017-08-02

Posted on August 03, 2017 at 01:31

Excellent work !

please reiterate to enlighten us all

'The correct answer i was enabled D and I cache at the same time with accelerator, and it cause this.' ( this and accelerator ??)

when you have the ART Accelerator ON, D-Cache ON I-Cache ON, then .. what doesn't work ?

megahercas6 · ‎2017-08-03

Posted on August 03, 2017 at 13:38

when you have the ART Accelerator ON, D-Cache ON I-Cache ON, then .. what doesn't work ?

my external 32b SRAM interface does not work, while at the same time, debugger shows that every thing is ok.

It just does not want to generate any hardware signals, same code, commented DI init functions, and viola, works like a charm

megahercas6 · ‎2017-08-03

Posted on August 03, 2017 at 14:06

Or to be more precise, i can't use //SCB_EnableDCache();, since it will stop my SRAM interface working.

Any ideas why it is happening?