K60 flexbus problems

[andrewclegg]

I'm trying to access a 16 bit fully multiplexed 64k CPLD which is hooked up to the external bus interface on chip select 0. I am using the TWR-K60F120M and MQX 4.1. The following code runs in a standard task.

#define A2D_CPLD_START_ADDRESS (*(volatile unsigned char*)(0xA0000000))
void flexbus_init(void) {
  SIM_SCGC7 |= SIM_SCGC7_FLEXBUS_MASK; // Enable the clock to the FlexBus
  SIM_CLKDIV1 |= SIM_CLKDIV1_OUTDIV3(0); // no divisor

  // Set the GPIO ports clocks
  SIM_SCGC5 = SIM_SCGC5_PORTB_MASK | SIM_SCGC5_PORTC_MASK | SIM_SCGC5_PORTD_MASK;

  // 16 bit multiplexing - so set 16 bit address, latch, then those 16 bits become data lines
  PORTB_PCR18 = PORT_PCR_MUX(5); // fb_ad[15]
  PORTC_PCR0 = PORT_PCR_MUX(5); // fb_ad[14]
  PORTC_PCR1 = PORT_PCR_MUX(5); // fb_ad[13]
  PORTC_PCR2 = PORT_PCR_MUX(5); // fb_ad[12]
  PORTC_PCR4 = PORT_PCR_MUX(5); // fb_ad[11]
  PORTC_PCR5 = PORT_PCR_MUX(5); // fb_ad[10]
  PORTC_PCR6 = PORT_PCR_MUX(5); // fb_ad[9]
  PORTC_PCR7 = PORT_PCR_MUX(5); // fb_ad[8]
  PORTC_PCR8 = PORT_PCR_MUX(5); // fb_ad[7]
  PORTC_PCR9 = PORT_PCR_MUX(5); // fb_ad[6]
  PORTC_PCR10 = PORT_PCR_MUX(5); // fb_ad[5]
  PORTD_PCR2 = PORT_PCR_MUX(5); // fb_ad[4]
  PORTD_PCR3 = PORT_PCR_MUX(5); // fb_ad[3]
  PORTD_PCR4 = PORT_PCR_MUX(5); // fb_ad[2]
  PORTD_PCR5 = PORT_PCR_MUX(5); // fb_ad[1]
  PORTD_PCR6 = PORT_PCR_MUX(5); // fb_ad[0]

  // Other lines - Output Enable, Read/Write, ChipSelect0, Address Latch Enable
  PORTB_PCR19 = PORT_PCR_MUX(5); // fb_oe_b [nEBI_OE]
  PORTC_PCR11 = PORT_PCR_MUX(5); // fb_rw_b [EBI_R/W]
  PORTD_PCR1 = PORT_PCR_MUX(5); // fb_cs0_b [nEBI_CSO]
  PORTD_PCR0 = PORT_PCR_MUX(5); // fb_ale   [EBI_ALE/nEBI_CSI]

  // Setup the chip select 1 control register (see K60P144M150SF3RM.pdf, section 33.3.3)
  FB_CSCR0 = FB_CSCR_PS(0x2) // 16-bit port
     //| FB_CSCR_BLS_MASK // data is right shifted, i.e. appears on bits 0-15
     | FB_CSCR_EXTS_MASK
  | FB_CSCR_AA_MASK // auto-acknowledge...
  | FB_CSCR_WS(0x2) // ...after 2 wait states
  | FB_CSCR_ASET(2) // assert chip select on 2nd clock edge after address
  ;

  // Chip select mask register
   FB_CSMR0 = FB_CSMR_BAM(0x0) //Set base address mask for 64K address space
   | FB_CSMR_V_MASK
   ;

  // Chip select address register
  FB_CSAR0 = (unsigned int)&A2D_CPLD_START_ADDRESS; //Set Base address
}

void flexbus_read_card_info() {
  volatile uint16_t raw_ID;
  _dcache_disable(); // see https://community.freescale.com/thread/306380
  raw_ID = (*(uint16_t*)(&A2D_CPLD_START_ADDRESS));
 printf("Raw ID: %d\n", raw_ID);
}

The task runs flexbus_init, then flexbus_read_card_info. Whatever I do, raw_ID seems to come out as 40960, i.e. 0xa000. The value should actually be 0x81 (which is stored in the CPLD).

Can anyone help me understand why this might be?

I also noticed in the bsp_twrk60f120m code, in init_hw.c it calls a function _bsp_flexbus_mram_setup which also sets up chip select 0; I have tried commenting out this function call but it makes no difference.

Many thanks!

[Hui_Ma]

Hi Andrew,

You are using Flexbus in multiplexed mode.

For the Flexbus 16bits data lines at FB_AD[15:0] pins, it need to set CSCRn[BLS] = 1, more detailed info, please check below picture:

And it need to check the hardware connection, for it using 16-bit data port, the CPLD address A0 pin should connect with Flexbus_AD[1] pin.

Wish it helps.
best regards
Ma Hui

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

[andrewclegg]

Thankyou very much for your quick reply. I forgot that I had commented out the BLS=1, thanks.

We currently have the CPLD Address[15:0] pins connected directly to AD[15:0] through an address latch. Looking in the reference manual K60P144M150SF3RM.pdf, section 33.4.1.2 says that the address and data lines should be connected from FB_AD0 upwards, however on the next paragraph it says that a 16 bit device connects to FB_AD[16:1]. Do you have any more information or an example schematic on why it starts from FB_AD1 and not FB_AD0? Should the FB_AD0 line be connected to anything?

Many thanks again

[andrewclegg · Accepted solution]

We spent the whole day with the logic analyzer and figured out the problem. I'll put the solution here incase someone else has the same problem.

Firstly it was OK to connect addr[15:0] directly to FB_AD[15:0], because our 16-bit CPLD does not use word addressing.

The problem was that in the ChipSelect Multiplexing Control Register (FB_CSPMCR), the group1 multiplex control was set to 1 (FB_CS1) rather than 0 (FB_ALE). This occurs in the MQX init code - hw_init.c, _bsp_flexbus_setup(). Because of this, the address latch enable was not working at all.

Thanks anyway for your help, you pushed us in the right direction.