Spansion SPI flash for XMOS Topic is solved

[Kasebe]

Hello everyone,

 

XMOS explains the use of M25P10-A as shown below link.

https://www.xmos.com/support/xtools/doc ... Flash%20Pr...

 

I'm going to use Spansion S25FL1-K.

http://www.spansion.com/Support/Datashe ... 1-K_00.pdf

I think we can use the SPI flash as well.

 

Does anyone have successful use of S25FL1-K?

[mon2]

Kasebe, we do not see any reason why this cannot be used to replace the original SPI Flash. Earlier this year we did a thorough review of SPI flash devices to determine the most suitable components for high speed data transfers involving QUAD SPI mode. For standard boot from SPI flash the boot loader uses the SPI standard and backwards compatible commands that are common to most (all ?) serial flash suppliers.

See here for a similar discussion:

http://beta.xcore.com/forum/viewtopic.p ... &view=next

From memory (pun intended), we found that Winbond devices (upto the time of the review) did not offer as strong pin drive strength as compared to Spansion. That is, in our lab, we applied the XMOS SliceKit with the DUT (device under test = Winbond for most of our tests then replaced with Spansion) + external wires which then mated with a Logic Analyzer (Zeroplus LAP-B 702000X). Using the XMOS xSoftIP to read/write to any of the target devices (any vendor Winbond, Spansion, Micron) with standard SPI modes worked correctly as the operating speeds are relatively slow. Shifting to QUAD SPI mode posed problems. We found numerous errors in the XMOS Github posted IP for Quad SPI mode which lead to weeks of rework. We validated our understanding by bringing in a dedicated QUAD SPI experimenter board supplied by Zeroplus to confirm our timing + opcodes to be correct. In hindsight a silly error on our part involved the length of our cables used for the debugging exercise = summary - keep the probes as short as possible. To resolve, we purchased heavily shielded logic analyzer probes from Zeroplus. However, a confirmed observation was that if Spansion is used under the same environment, the code would work correctly as compared to Winbond - even with the longer and un-shielded cabling. We alerted Winbond on this issue and they then supplied us more current versions of the SPI flash devices with their latest die - the datasheets for these new versions reports new registers for PIN DRIVE STRENGTH. If you are applying onto a PCB with a standard and clean layout then very doubtful if you will see the same issues. By default, from our review, Spansion was able to offer the QUAD SPI mode and of course the backwards compatible standard SPI opcodes + offers a strong pin drive strength = "out of the box". We have not yet reviewed the latest supplied Winbond devices but Winbond now features even more QUAD commands (unique to specific Winbond devices only) that will kick up the transfer modes if the IP is applied.

It will be best for you to test a few of the referenced SPI devices but if it helps, we can check the lab for the same P/N. Are you testing on SliceKit or startKit ? We can replace the SMD device with Spansion and test for the default boot loader R/W access.

One important discovery during our review was the support of the narrow (150 mil) vs wide (208 mil) 8 pin SOIC packages across vendors. For your PCB layout, consider a unified PCB landing pattern (that is, nest the narrow 150 mil pattern inside the 208 mil) so you can substitute at assembly time for either package. In a pinch, it may help you during a parts shortage.

 

[Kasebe]

Hi mon2,

 

Thank you for your kind explanation.

I'm just considring SPI flash devices, so the SPI flash has not been implemented yet.

The reason why is easy for us to be supplied the Spansion.

I understood we must use original SPI for stable development.

 

Regards,

Kasebe