LAUNCHXL-F28379D: Question about EMIF1 - 16-bit 256Mb SDRAM

Yang Lei

Part Number: LAUNCHXL-F28379D
Other Parts Discussed in Thread: C2000WARE

Hello. I'm using LAUNCHXL-F28379D and I built a small memory board that can connect to LAUNCHXL-F28379D through J9. After I connect the memory board to LaunchPad, I did a quick memory write/read test and it gives me weird results. All the even address memory is not working correctly. It feels like this issue is caused by address line GPIO38/EM1A0. Then, I did two tests as follow:

1. I write to 0x8000 0000 and 0x8000 0001, I expected A0 should be different in these two cases, but they are both low.

2. I configure GPIO38 as an output GPIO and output high, I can measure high on the memory chip. This proves the connection is Ok.

Can someone please help me with this issue? I attached my main.c source file and the memory browser screenshot. Thanks.

Best regards.

Yang

/**
* main.c
*/

#include "F28x_Project.h"

extern void LaunchXL_F28379D_CPU1_Init(void);

volatile Uint16 *emif_pointer = (Uint16 *)0x80000000;
Uint32 emif_bias = 0;
Uint16 write_data = 0;
Uint16 read_data = 0;
Uint32 failed_count = 0;

void main(void)
{
InitSysCtrl();
LaunchXL_F28379D_CPU1_Init();

PieCtrlRegs.PIECTRL.bit.ENPIE = 1; // Enable the PIE block
IER = M_INT1 | M_INT3 | M_INT9;
EINT;

while(1)
{
failed_count = 0;
write_data = 0;
write_read = 0;
for(emif_bias=0;emif_bias<0x10000000;emif_bias++)
{
*(emif_pointer+emif_bias) = write_data;
write_data++;
}
write_data=0;
write_read = 1;
for(emif_bias=0;emif_bias<0x10000000;emif_bias++)
{
read_data = *(emif_pointer+emif_bias);
if(read_data!=write_data)
failed_count++;
write_data++;
}
}
}

over 4 years ago

0 Hareesh Janakiraman over 4 years ago

TI__Guru* 95315 points

Yang,

Could you please attach the wiring diagram between the boards?

0 Hareesh Janakiraman over 4 years ago in reply to Hareesh Janakiraman

TI__Guru* 95315 points

Also, have you taken a look at www.ti.com/lit/sprac96?

Another helpful resource is the C2000-EMIF_ConfigurationTool.xlsx tool in C:\ti\c2000\C2000Ware_4_00_00_00\boards\TIDesigns\F28379D_EMIF_DC

0 Vivek Singh over 4 years ago in reply to Hareesh Janakiraman

TI__Guru** 117431 points

Yang,

If you are using 16bit SDRAM memory, make sure you are configuring NM bit in SDRAM_CR register to '1'. By default it's '0'.

Regards,

Vivek Singh

0 Yang Lei over 4 years ago in reply to Hareesh Janakiraman

Prodigy 80 points

Hi Hareesh,

Thanks for your reply. Please check the attached wiring diagram. The left side is the memory chip and the right side is the J9 connector.

I used this diagram before and have no issue. I don't know why it went into this issue when I use it for LaunchPad.

Thanks!

0 Yang Lei over 4 years ago in reply to Vivek Singh

Prodigy 80 points

Hi Vivek,

Thanks for your reply. I just checked the register and NM bit is "1". I followed the EMIF configuration in TI's example "emif1_16bit_sdram_far.c".

0 Hareesh Janakiraman over 4 years ago in reply to Yang Lei

TI__Guru* 95315 points

Yang,

Please expect some delay due to the holidays. Many in our team are on vacation. We will work on getting you an answer before the end of next week.

0 Chester Gillon over 4 years ago

Guru 92251 points

Yang Lei said:
Hello. I'm using LAUNCHXL-F28379D and I built a small memory board that can connect to LAUNCHXL-F28379D through J9.

Which revision of LAUNCHXL-F28379D are you using?

LAUNCHXL-F28379D overview user's guide (Rev. C) notes that depending upon the revision of the launchpad you made need to change jumpers, or move zero ohm resistors, to route all the SDRAM EMIF signals to J9:

And the schematic section which shows the zero ohm routing resistors on the revision 2 launchpad:

0 Yang Lei over 4 years ago in reply to Chester Gillon

Prodigy 80 points

Hi Chester. I'm using Ver 2.0. I double-checked the 0 Ohm signal routing resistor and it is correct. I've already routed all the necessary signals to J9. Also, GPIO38 doesn't have 0 Ohm resistor. Also, as I mentioned before, I did a test that I configure GPIO38 as output and checked the connection between F28379D chip to the SDRAM chip. So, I don't think it is a connection issue. Thanks!

Regards,

Yang

0 Hareesh Janakiraman over 4 years ago in reply to Yang Lei

TI__Guru* 95315 points

DQMH & DQML is reversed for U1. Is this intentional?

0 Yang Lei over 4 years ago in reply to Hareesh Janakiraman

Prodigy 80 points

Hi Hareesh,

Thanks for your reply. Yes, I swapped the high 8-bit and low 8-bit for the data bus so I also swapped DQMH and DQML. But after I read sprac96, it says DQMH and DQML are not used if the data bus width is 16-bit. So, could this be the issue? Thanks!

Regards,

Yang

0 Hareesh Janakiraman over 4 years ago in reply to Yang Lei

TI__Guru* 95315 points

I presume you are referring to the below:

Dynamic control of the byte enable signals is not required for 16-
bit memories because the C28x is 16-bit word addressable. The
memory byte enable signals can be tied low (permanently
enabled) for 16-bit memories. Byte enable signals for 32-bit
memories can also be tied low if 16-bit writes and reads are not
used.

Please tie them low as suggested and see if that fixes the issue.

0 Yang Lei over 4 years ago in reply to Hareesh Janakiraman

Prodigy 80 points

Hi Hareesh,

Thanks for your support. Unfortunately, this doesn't solve my problem. I tied DQMH and DQML to the nearest GND pin, but the issue still exists.

Regards,

Yang

0 Hareesh Janakiraman over 4 years ago in reply to Yang Lei

TI__Guru* 95315 points

Please follow the same wiring you find in C:\ti\c2000\C2000Ware_4_00_00_00\boards\TIDesigns\F28379D_EMIF_DC\F28379D_EMIF_DC_SCH.pdf and try an example from C:\ti\c2000\C2000Ware_4_00_00_00\driverlib\f2837xd\examples\cpu1\emif and let us know what you observe.

0 Yang Lei over 4 years ago in reply to Hareesh Janakiraman

Prodigy 80 points

Hi Hareesh,

I built a new board and I fixed DQMH and DQML swap. But I still swapped some bits for layout simplicity. Unfortunately. the issue still exists.

I also tried C:\ti\c2000\C2000Ware_4_00_00_00\driverlib\f2837xd\examples\cpu1\emif\emif_ex4_16bit_sdram_far_lpm.c. The variable "testStatusGlobal" returns "0xDEADDEAD" and "errCountGlobal" returns "2".

Any comments? Thanks!

Best regards,

Yang

0 Hareesh Janakiraman over 4 years ago in reply to Yang Lei

TI__Guru* 95315 points

Yang,

The C2000ware example is a tested example. If it doesn't work as intended, it points to some hardware issue on your end. Can you share your complete schematics with me privately? You can do this after initiating a friendship request with me.

+1 Yang Lei over 4 years ago in reply to Hareesh Janakiraman

Prodigy 80 points

Hi Hareesh,

I figured this out. The hardware has no issue. The issue is that I didn't set GPIO Qualifier Select correctly. I found the function "setup_emif1_pinmux_sdram_16bit(0)" works well but my own code doesn't work. Then, I compared the code and found out the data line should use asynchronous input qualification. Thanks for your support!

Regards,

Yang

C2000™︎ microcontrollers

C2000 microcontrollers forum

LAUNCHXL-F28379D: Question about EMIF1 - 16-bit 256Mb SDRAM