QSPI - Read UNIQUE ID from W25Qxx NOR memory

[uilter]

Hi ST community!

I'm a liltle confused about how to read the UNIQUE ID from memory NOR W25Qxx. In my case a use the W25Q128 - https://www.winbond.com/hq/product/code-storage-flash-memory/serial-nor-flash/?__locale=en&partNo=W25Q128JV

We already use this memory in another custom boards.
We able to flash, create external loaders and use it with touchgfx.

This custom board has a STM32H753.

I want to understand the correct way to read this instruction UNIQUE ID (4Bh). To get this value, the datasheet says after the INSTRUCTION (4Bh) I need to add 4 bytes of DUMMY clock (32 clocks). After this, I will able to receive the 64 bits from the UNIQUE ID.

I'm a little confused about my tests and would like some help from the community to know if some of the two paths I show below are correct or none of them haha

1) My first try was using the code below, but I always get corrupt data.
- I put the max dummy cycles allowed (31). Check it on the HAL lib and reference manual.
- read 8 bytes (64 bits from the UNIQUE ID)

Looking the logic analyzer, I figure out the data is shifted to left. The last byte was not readed.

2) The second try was the code bellow:
- I sum the 8 bytes from UNIQUE ID with the Dummy bytes. 
- the first 4 bytes are zero and the UNIQUE ID start at index 4 in pData array.

Looking the logic analyzer, the data is fine.

Please, anyone can advise me the correct way to read this UNIQUE ID from W25Q128 NOR memory?

Many thanks!

[FBL]

Hello @uilter,

Some updates in the configuration should be adjusted:

s_Command.DummyCycles = 4; //according to the datasheet

s_Command.NbData      = 8; // 64bits

If still facing issues, please keep me updated!
 

[JJhin.1]

Im also facing QUADSPI issue with touchGFX... in my case TouchGFX Background image is getting saved into QSPI memory... im also using W25Q128 ic for storing ... but whatever data is getting saved on QSPI is not getting displayed on Display.. but Those data who are getting saved into flash memory is getting displayed properly.

and more thing is ... when i used STemwin for same configuration every thing is getting displayed properly :thinking_face:

[raptorhal2]

It would be much clearer if everyone used s_command.DummyBytes.

[raptorhal2]

And if s_command.NbData was s_command.NBytesData.

And if I could edit my post.

[uilter · Accepted solution]

We did tests in 3 different boads. All custom boards has this memory (W25Q128xx).
The difference between this board is the MCU: STM32H753BI, 753II and 753ZI.

The following image show the code to read this UNIQUE ID and bellow the logic analyzer.
You can see the debug mode, the array pData match with the logic analyzer.

We setup the logic analyzer adding time markers and a counter in CLOCK line of the QSPI.
According the datasheet, we need in total 104 clocks, splitted in:
- A marker, 8 clocks: instruction (4Bh)
- B marker, 32 clocks: dummy cycles
- C marker, 64 clocks: 64-bits UNIQUE ID
- D marker, total of 104 clocks. This match with Figure 42 wave form in W25Q128 datasheet.

Whatever is the value for variable s_command.DummyCycles, I not able to see a different behaviour on my logic analyzer even in the oscilloscope...

We test it on 3 different custom boards, as I mentioned. The MSB byte 0xE4 always repeat. The rest of pData value is different. Putting the logic analyzer on it, the value matches with Data.

[uilter]

Hi @FBL , thanks for your reply!

We test according your suggestion, but not works. This is my output.

[FBL]

Hello again @uilter 

Could you keep the receive buffer size 8 and empty uint8_t pData[8] = {}.

[uilter]

Hi @FBL , this is the result:

[Tesla DeLorean]

How's that going to work? It's frankly far easier to make the buffer bigger, keep the dummy cycles within bounds, and simply discard the leading slack bytes to extract the unique identifier.

Winbond doesn't have to cover the whole 64-bit number space, just provide something that's part unique across their offerings, this may mean a lot of static fields, reporting batch, wafer, and die position, so is less reliant on a huge database and issuing test-sets an audited / traced range of numbers (say like MAC addresses) at every run or test fixture..