CCS/TMS320F28335: F28335 write data to fpga through xintf, but the data is incorrect

fan fanghao

Tool/software: Code Composer Studio

Hi all, I got a very weird issue on 28335 xintf interface right now. In my board, I connected low data bus(bit15~0) xintf pins to a fpga, and I configured xintf to 16 bit bus mode. The issue is that when dsp write data to fpga,the data on bus is not the same as dsp writed,you see this in picture one, but read operation is ok.

EALLOW;

XintfRegs.XINTCNF2.bit.XTIMCLK = 1;
// No write buffering
XintfRegs.XINTCNF2.bit.WRBUFF = 0;
// XCLKOUT is enabled
XintfRegs.XINTCNF2.bit.CLKOFF = 0;
// XCLKOUT = XTIMCLK/2
XintfRegs.XINTCNF2.bit.CLKMODE = 0;

XintfRegs.XTIMING7.bit.XWRLEAD = 1;
XintfRegs.XTIMING7.bit.XWRACTIVE = 7;
XintfRegs.XTIMING7.bit.XWRTRAIL = 1;
// Zone read timing
XintfRegs.XTIMING7.bit.XRDLEAD = 3;
XintfRegs.XTIMING7.bit.XRDACTIVE = 7;
XintfRegs.XTIMING7.bit.XRDTRAIL = 3;

// double all Zone read/write lead/active/trail timing

XintfRegs.XTIMING7.bit.X2TIMING = 0;

// Zone will sample XREADY signal

XintfRegs.XTIMING7.bit.USEREADY = 0;
XintfRegs.XTIMING7.bit.READYMODE = 0; // sample asynchronous

// Size must be either:
// 0,1 = x32 or
// 1,1 = x16 other values are reserved
XintfRegs.XTIMING7.bit.XSIZE = 3;

// Bank switching
// Assume Zone 7 is slow, so add additional BCYC cycles
// when ever switching from Zone 7 to another Zone.
// This will help avoid bus contention.
XintfRegs.XBANK.bit.BANK = 7;
XintfRegs.XBANK.bit.BCYC = 7;

///GPIO config

GpioCtrlRegs.GPCMUX1.bit.GPIO64 = 3; // XD15
GpioCtrlRegs.GPCMUX1.bit.GPIO65 = 3; // XD14
GpioCtrlRegs.GPCMUX1.bit.GPIO66 = 3; // XD13
GpioCtrlRegs.GPCMUX1.bit.GPIO67 = 3; // XD12
GpioCtrlRegs.GPCMUX1.bit.GPIO68 = 3; // XD11
GpioCtrlRegs.GPCMUX1.bit.GPIO69 = 3; // XD10
GpioCtrlRegs.GPCMUX1.bit.GPIO70 = 3; // XD19
GpioCtrlRegs.GPCMUX1.bit.GPIO71 = 3; // XD8
GpioCtrlRegs.GPCMUX1.bit.GPIO72 = 3; // XD7
GpioCtrlRegs.GPCMUX1.bit.GPIO73 = 3; // XD6
GpioCtrlRegs.GPCMUX1.bit.GPIO74 = 3; // XD5
GpioCtrlRegs.GPCMUX1.bit.GPIO75 = 3; // XD4
GpioCtrlRegs.GPCMUX1.bit.GPIO76 = 3; // XD3
GpioCtrlRegs.GPCMUX1.bit.GPIO77 = 3; // XD2
GpioCtrlRegs.GPCMUX1.bit.GPIO78 = 3; // XD1
GpioCtrlRegs.GPCMUX1.bit.GPIO79 = 3; // XD0

GpioCtrlRegs.GPBMUX1.bit.GPIO40 = 3; // XA0/XWE1n
GpioCtrlRegs.GPBMUX1.bit.GPIO41 = 3; // XA1
GpioCtrlRegs.GPBMUX1.bit.GPIO42 = 3; // XA2
GpioCtrlRegs.GPBMUX1.bit.GPIO43 = 3; // XA3
GpioCtrlRegs.GPBMUX1.bit.GPIO44 = 3; // XA4
GpioCtrlRegs.GPBMUX1.bit.GPIO45 = 3; // XA5
GpioCtrlRegs.GPBMUX1.bit.GPIO46 = 3; // XA6
GpioCtrlRegs.GPBMUX1.bit.GPIO47 = 3; // XA7

GpioCtrlRegs.GPCMUX2.bit.GPIO80 = 3; // XA8
GpioCtrlRegs.GPCMUX2.bit.GPIO81 = 3; // XA9
GpioCtrlRegs.GPCMUX2.bit.GPIO82 = 3; // XA10
GpioCtrlRegs.GPCMUX2.bit.GPIO83 = 3; // XA11
GpioCtrlRegs.GPCMUX2.bit.GPIO84 = 3; // XA12
GpioCtrlRegs.GPCMUX2.bit.GPIO85 = 3; // XA13
GpioCtrlRegs.GPCMUX2.bit.GPIO86 = 3; // XA14
GpioCtrlRegs.GPCMUX2.bit.GPIO87 = 3; // XA15
GpioCtrlRegs.GPBMUX1.bit.GPIO39 = 3; // XA16
GpioCtrlRegs.GPAMUX2.bit.GPIO31 = 3; // XA17
GpioCtrlRegs.GPAMUX2.bit.GPIO30 = 3; // XA18
GpioCtrlRegs.GPAMUX2.bit.GPIO29 = 3; // XA19

GpioCtrlRegs.GPBMUX1.bit.GPIO34 = 3; // XREADY
GpioCtrlRegs.GPBMUX1.bit.GPIO35 = 3; // XRNW
GpioCtrlRegs.GPBMUX1.bit.GPIO38 = 3; // XWE0

GpioCtrlRegs.GPBMUX1.bit.GPIO36 = 3; // XZCS0
GpioCtrlRegs.GPBMUX1.bit.GPIO37 = 3; // XZCS7
GpioCtrlRegs.GPAMUX2.bit.GPIO28 = 3; // XZCS6

EDIS;

asm(" RPT #7 || NOP");

the test code is shown as below:

Uint16 *ExRamStart = (Uint16 *)0x200000;

*(ExRamStart + 1) = 0x5a5a;
*(ExRamStart + 2) = 0x1234;
*(ExRamStart + 3) = 0xa5a5;

Some bits of the data is not correct, this really confuse me.

By the way, I'm using zone7 for fpga memory space, and all dsp read function works good.

Does any TI experts or other experts can help me? Thanks in advance.

over 5 years ago

0 KARTHIKEYAN RAJAMANICKAM over 5 years ago

TI__Genius 11645 points

Hi,

One of our experts will get back to you by Tuesday, since Monday is a US holiday.

0 fan fanghao over 5 years ago in reply to KARTHIKEYAN RAJAMANICKAM

Prodigy 20 points

Hi Karthik,

Thanks for your reply,there are some new situations:

1. When the initialization finished, before the dsp didn't write no data to the space, but the address 0x20000~0x200fff has filled with datas such as 0x3ff, 0x8bf etc,where did these data come from ?

2. When dsp write data to the address use *(0x201100) = 0x5a5a,the low 10bits is all 1,that means the data is 0x5dff; But when i write 0x5a5a in memory at the same address, the data on bus is correct.

Best Regards,

Fanghao.

0 tlee over 5 years ago in reply to fan fanghao

TI__Guru 62975 points

fan fanghao said:
By the way, I'm using zone7 for fpga memory space, and all dsp read function works good.

How are you able to determine that reads are working?

fan fanghao said:
When the initialization finished, before the dsp didn't write no data to the space, but the address 0x20000~0x200fff has filled with datas such as 0x3ff, 0x8bf etc,where did these data come from ?

If the locations have not been written to, the contents are invalid, uninitialized values. In the case where you are interfacing with an FPGA, it is determined by the FPGA.

fan fanghao said:
When dsp write data to the address use *(0x201100) = 0x5a5a,the low 10bits is all 1,that means the data is 0x5dff; But when i write 0x5a5a in memory at the same address, the data on bus is correct.

Please clarify how you perform write 0x5a5a in memory at the same address and how this is different from *(0x201100) = 0x5a5a. When you say that all lower 10-bits are 1, are you observing the voltage with an oscilloscope? This would be the preferred method because it can show you if there is contention on the bus.

C2000™︎ microcontrollers

C2000 microcontrollers forum

CCS/TMS320F28335: F28335 write data to fpga through xintf, but the data is incorrect