TMS320VC5509A: TMS320C5509A EMIF 16 bit External SDRAM (CE1) Memory Access issue.

Thanks Ming. I tried both address spaces ( Data Space and Program Space) same behavior. If its program it fails at 0x403746@PROGRAM and fails at 0x203746@DATA if data space.

Hi Ming, I could able to view the memory area in the debug mode.

Ramesh

Hi Ramesh,

You said in your previous post:

3. BOOT mode set to (16bit  Async memory (on CE1 space) or Execute from 16bit  Async memory (on CE1 space) both didnt work.

Can you explain to me how you initialized your board and set up the EMIF? Are you using GEL file or your application code? Are you using JTAG to load your code or boot from flash or EEPROM?

Best regards,

Ming

Hi Ming,

Hardware BOOTM GPIO pins connected to DSP as (GPIO0=1, GPIO1=1,GPIO2=1,GPIO3=0). Do I need to configure these GPIO's in GEL file as well?..  Initially I tried EMIF initialization with the application code as below but it didn't work so I'm using the GEL file to initialize the EMIF . Also loading the software via JTAG. 

Please find EMIF Configuration details below:-
EBSR_REG = 0x0001;
EMIF_RST_REG = 0xFFFF;
EMIF_EGCR_REG = 0x0221;
EMIF_CE1_1_REG = 0x3FFF;
EMIF_CE1_2_REG = 0xFFFF;
EMIF_CE1_3_REG = 0x00FF;
EMIF_SDC1_REG = 0x2911;
EMIF_SDPER_REG = 0x0410;
EMIF_INIT_REG = 0x07FF;

Thanks,

Ramesh

Hi Ramesh,

The BootM pins are read when the DSP is brought out of reset. They will determine the ROM code execution. To get the BootM pins in GEL file is too late. The EMIF setting should be done in GEL file or by ROM code, because you will need the correct EMIF setting before loading (JTAG or ROM code) the application code into external memory.

Since you are able to access the external memory after JTAG connection and before loading the program, it means the EMIF setting is correct. What the JTAG do is simply copy the data and program from the OUT file to the external memory section by section. It will verify the content just loaded. If there is any mismatching, it will report validation error. I do not know why this error happens.

Can you check on the data section which is located at the address reported validation error (0x403746@PROGRAM or 0x203746@DATA). Or you can send me the MAP file for your application. I will check for you.

BTW, what is the memory model are you using, small, large or huge?

Best regards,

Ming  

Hi Ming,

As per .map file (0x400000) starts at .pinit, .twiddle, cinit and then .const, .switch, .text . Looks like JTAG copies the .pinit, .twiddle, cinit , .cons, .switch but exactly at .text section (i.e. 0x403746 ) validation error occurred. 

Memory model is Large. 

Thanks,

Ramesh 

Hi Ming,

The .text size is huge and it wont fit into the internal memory. So I tried configuring the .text in the data space 16bit (0x200000) of external memory (CE1). Getting validation error at 0x203746@DATA. Boot mode is Parallel EMIF-Boot Mode.

Thanks,

Ramesh

Hi Ramesh,

Can you try the same with simplified application code, so that the .text can be fit in internal memory?

Best regards,

Ming

Hi Ming,

I tried moving DATA section to the external memory ( 0X200000) and pointed code to the internal memory but throwing below errors and trying to understand why VECs section allocation issue occurred.

error: can't allocate .vectors, size 00000100 (page 0) in VECS (avail:
>> Compilation failure
makefile:186: recipe for target 'Main_App.out' failed
00000000)

Memory Allocation:
 VECS : origin = 0x000900, length = 0x000100
 CODE : origin = 0x000A00, length = 0x03F600 /* Internal Memory */
 DATA : origin = 0x200000, length = 0x020000 /* External memory */

Hi Ramesh,

Can you allocate VECS bigger size like 0x200?

Best regards,

Ming

Hi Ming,

I adjusted other section space and VECS error is no longer exist but now getting below error (i.e. I tried Data section origin 0x400000 same error but address location is 0x20495E@DATA.)

File Loader: Verification failed: Values at address 0x10495E@DATA do not match Please verify target memory and memory map.

Thanks,

Ramesh G

Hi Ramesh,

0x10495E@DATA seems to be incorrect. Your data area should start at 0x200000@DATA right?

Best regards,

Ming

Hi Ming,

Yes, data should start at 0x200000 for some reason its divided by 2 and throws error at 0x10495E. If I change it to data origin as 0x400000 then error throws at 0x20495E.

Thanks,

Ramesh 

Hi Ramesh,

Can you access (write then read) to the external memory at or beyond 0x20495E using CCS memory browser?

Best regards,

Ming

Hi Ming,

Yes, I could able to view 0x20495E in the memory browser.

Thanks,

Ramesh G

Hi Ramesh,

Can you write something other than zero to this area using CCS memory browser?

Best regards,

Ming

Hi Ming,

I could able to write it program space at 0x40495E and it does reflect in the data space 0x20495E. Also it writes into entire memory.

Thanks,

Ramesh G

Hi Ramesh,

Can you share your MAP file?

Best regards,

Ming

Hi Ramesh,

In your map file:

 [ 004092b8 ]  0020495c          *   0000194b   FftOperations.obj (.bss)

 0020495c    _finalNormQ4

It indicates that the data buffer finalNormQ4 is where the problem is. What is the data type of finalNormQ4? 8-bit, 16-bit, 32-bit or s structure?

If it is 8-bit or structure with 8-bit variables, then it is the problem.

Best regards,

Ming

Hi Ming,

The data type of finalNormQ4 variable is Int16 not a 8-bit or structure with 8-bit variable.

Thanks,

Ramesh G

Hi Ramesh,

Sorry for sending the incorrect variable to check. It should be the 

0020495e    _AltInternalDelayQ15 

What is the data type of AltInternalDelayQ15?

Best regards,

Ming

Hi Ming,

No worries. The data type of AltInternalDelayQ15 variable is Uint32.

Thanks,

Ramesh G

Hi Ming,

CCS throwed below error after setting verification option to "No Verification".

Trouble Writing Memory Block at 0x205a56 on Page 1 of Length 0x9: (Error -1144 @ 0x205A5A) Device core is hung. The debugger attempted to recover debug control, but was unsuccessful. Power-cycle the board. If error persists, confirm configuration and/or try more reliable JTAG settings (e.g. lower TCLK). (Emulation package 6.0.407.3) 
File Loader: Verification failed: Target failed to write 0x205A56@DATA

Thanks,

Ramesh

Hi Ramesh,

I cannot find the variable at 0x205A56@DATA in your map file. What is the data type of the variable?

Best regards,

Ming

Hi Ming,

Yes, I dont see any variable at 0x205A56 in the map file and I'm not sure why CCS throws error at 0x205A56 location.

Thanks,

Ramesh

Hi Ming,

I could able to write/read to the 0x205A56 and 205A5A memory locations.

Thanks,

Ramesh

Hi Ming,

JTAG default speed selected 35 MHz.  Also, tried ranges between 0.5 to 35Mhz still didn't work.

Thanks,
Ramesh

Hi Ramesh,

Are you using the onboard JTAG or an external JTAG? What is the JTAG you are using to load the program XDS510,or XDS560?

Best regards,

Ming

Hi Ming,

We are using USB 560 JTAG EMULATOR ( BH-USB-560BP). 

Thanks.

Ramesh G.

Hi Ming,

We installed the same external SD RAM into the DSK Board and we could able to load the software without any issue (External CE1) and not sure why in our actual board its not working. 

https://www.digikey.in/htmldatasheets/production/1288796/0/0/1/tms320vc5509a-dsk-tech-ref.html

Thanks,

Ramesh G

Hi Ming,

Yes, we did compare both ( DSK and LRU Card) pin connection wiring and both are identical. 

Thanks,

Ramesh G

Hi Ramesh,

Can you try the following in the GEL for your custom board:

hotmenu C5509_EVM_EMIF_Init()
{
*(unsigned int*)0x6C00@IO=0x0001; /* EXBUSSEL */
*(unsigned int*)0x0801@IO=0xFFFF; /* EMI_RST */
*(unsigned int*)0x0800@IO=0x0221; /* EGCR */
*(unsigned int*)0x0803@IO=0x3FFF; /* EMIF CE0_1 */
*(unsigned int*)0x0804@IO=0x5FFF; /* EMIF CE0_2 */
*(unsigned int*)0x0805@IO=0x5FFF; /* EMIF CE0_3 */
*(unsigned int*)0x0806@IO=0x122A; /* EMIF_CE1_1 */
*(unsigned int*)0x080F@IO=0x5938; /* SDC1 */
*(unsigned int*)0x0810@IO=0x091D; /* SDPER */
*(unsigned int*)0x0811@IO=0x0FFF; /* SDCNT */
*(unsigned int*)0x0813@IO=0x035F; /* SDC2 */
*(unsigned int*)0x0812@IO=0x0000; /* INIT */
}

Bet regards,

Ming

Hi Ming,

DSK board works fine with the above GEL but address should be divide by 2 instead of 0x200000.  Could able to write  in 0x100000 because of data space?. Also we tried the same gel file init script in the final LRU board with CE1 configuration it didnt work 0x100000 or 0x200000.

Thanks,
Ramesh G.

Hi Ramesh, 

My colleague I mentioned yesterday is in training this week. I will talk to him early next week and get back to you as soon as I hear anything.

Best regards,

Ming

Hi Ramesh,

I talked to my colleague today. Here are a few things you can do:

1. In C5509A DSK, there is a dip switch S3-1 to control the value of GP0

Which in turn determines the A0-13 used for HPI or EMIF. Make sure it is used for EMIF addressing on your board.

2. The other thing you can try is to switch SDRAM to CE0 on your board or CE1 on C5509A DSK to confirm the CE1 can be used for SDRADM.

Unfortunately, other than the two suggestions above, we do not have any more suggestions.  

Best regards,

Ming