LP-AM261: AM261x LaunchPad: Changing OSPI from Octal to Single/Quad Mode

Hello,

Thank you for reaching out with your detailed questions about OSPI flash configuration on the AM261x LaunchPad. Before we dive into troubleshooting the PHY mode and Single/Quad mode issues you're experiencing, I need to gather some essential information about your setup to provide you with accurate guidance.

First, can you please confirm which specific flash part number you're using on your custom board? 

Second, did you update your syscfg for the respective flash part and the mode configuration?

Best regards,
Mayank Shadwani

Hello Mayank,

Thank you for your response.

Here is the information you requested:

1. Flash device part number

• On the AM261x LaunchPad: MX25UW064 (the actual OSPI flash mounted on the board)
• On my custom board: ISSI IS25LP128F

If needed, I can provide the full datasheet or relevant timing parameters.

2. SysConfig updates
Yes, I updated the SysConfig settings to match the flash device and the desired mode (Single / Quad).
Specifically:

• Set the OSPI protocol to 1S-1S-1S (or 1S-1S-4S when testing Quad).
• Disabled PHY mode when testing Single/Quad.
• Adjusted device-specific parameters such as page size, block size, dummy cycles, JEDEC ID, etc., as applicable.

If there are additional SysConfig fields that must be modified for Single/Quad operation on the AM261x, I would appreciate your guidance.

Best regards,
Akira Higuchi

Hi,

Thank you for the advice.
As you suggested, after setting the 4-Byte Addressing Enable Sequence to 0x21, I was able to confirm that read and write operations work correctly in 1S-1S-1S mode.
It looks like I can now proceed to the next verification step.

Thanks again for your support.

Best regards,
Higuchi

Hi,

Thank you for the guidance.
Using a custom board, I have already obtained the JSON file generated by the OSPI diagnostic example.
If needed, I can attach the JSON file here for your reference.

When running QSPI on the EVM, besides setting the 4-Byte Addressing Enable Sequence to 0x21, are there any other settings that need to be configured?

Thanks again for your support.

Hi,

Thank you for the guidance.
Using a custom board, I have already obtained the JSON file generated by the OSPI diagnostic example.
If needed, I can attach the JSON file here for your reference.

When running QSPI on the EVM, besides setting the 4-Byte Addressing Enable Sequence to 0x21, are there any other settings that need to be configured?

Thanks again for your support.

Akira Higuchi said:

When running QSPI on the EVM, besides setting the 4-Byte Addressing Enable Sequence to 0x21, are there any other settings that need to be configured?

May I know how QSPI protocol is running on the EVM, the EVM has a macronix flash which does not have QSPI support.

When running on custom board, you would need to change the boot mode and then align the flash configurations by uploading the JSON file

Akira Higuchi said:

besides setting the 4-Byte Addressing Enable Sequence to 0x21

This is done is 4 byte addressing is required or is supported by the flash. 

Regards,

Aswin

Hi Aswin,

Thank you very much for the clarification.
I understand that the EVM does not support QSPI due to the Macronix flash device.

Just for verification, I tried applying the Single-mode settings we discussed to the SBL project on the EVM.
However, the SBL did not run correctly after applying these changes.

Could you please confirm whether the SBL supports Single mode operation in the first place?
I would like to understand if the issue is due to my configuration or if SBL inherently does not support this mode.

Thanks again for your support.

Best regards,
Higuchi

Hello,

The SBL and Application are using the same ospi and flash driver. SBL can also work in 1s-1s-1s mode. There is no restrictons. The only thing that can be different is the flash confogurations in sysconfig. Can you please check the flash confogurations between the SBL and the working application side by side and check if those are same.

Regards,

Aswin

Hello Aswin,

I compared the flash configurations between the SBL and the application, and they appear to be identical.

While debugging, I found that when running the SBL in 1s-1s-1s mode, the following issue occurs:

Inside the SBL, the function below is called:

status = Bootloader_parseMultiCoreAppImage(bootHandle, &bootImageInfo);

Within this function, the magic number is checked here:

status = config->fxns->imgReadFxn(&mHdrStr, sizeof(Bootloader_MetaHeaderStart), config->args);

if (mHdrStr.magicStr != BOOTLOADER_META_HDR_MAGIC_STR)
{
status = SystemP_FAILURE;
}

The expected value of mHdrStr.magicStr is 0x5254534D, but the actual value being read is 0x54524D53, which results in a failure.

This seems to indicate that the data read from flash is being shifted or interpreted incorrectly only in 1s-1s-1s mode.

Do you have any suggestions on what might cause this behavior?

Regards,
Higuchi

Hello,

Let me try this on my side.

Regards,

Aswin

Thank you, Aswin.
Please let me know if I can provide any further details.
I appreciate your support.

Regards,
Higuchi

Hello, 

I was able to observe the issue. The ELF header is being read incorrectly during the image parsing procedure. Let me try to resolve and issue and get back to you as soon as possible.

Regards,

Aswin

Hello Aswin,

The issue has been resolved.
As a result, we are now able to boot the SBL in QSPI mode on our custom board and successfully launch the application.

Thank you very much for your kind support and cooperation.

Best regards,
Higuchi

Hello, 

Why SBL is not working in 1s-1s-1s mode is because,

The macronix flash which is on LP has a property that, when writing in 8D-8D-8D mode, it swaps the bytes and will write it. Basically the write does not happen the same way in 1s-1s-1s mode and 8d-8d-8d mode. So when we are flashing, the data is written in 8d-8d-8d mode and the SBL tries to read in 1s-1s-1s mode. So for SBL to work in 1s-1s-1s mode with the macronix flash which is on LP, the flasher used in uniflash also should be in 1s-1s-1s mode.

Regards,

Aswin

Hello, 

Why SBL is not working in 1s-1s-1s mode is because,

The macronix flash which is on LP has a property that, when writing in 8D-8D-8D mode, it swaps the bytes and will write it. Basically the write does not happen the same way in 1s-1s-1s mode and 8d-8d-8d mode. So when we are flashing, the data is written in 8d-8d-8d mode and the SBL tries to read in 1s-1s-1s mode. So for SBL to work in 1s-1s-1s mode with the macronix flash which is on LP, the flasher used in uniflash also should be in 1s-1s-1s mode.

Regards,

Aswin