HyperRAM stability issue with 528MHz core

[p_shep]

Hello,

We just changed our RT1064 part from the 600MHz variety to the 528MHz one.

On the 600MHz part we had HyperRAM running happily at 216MHz for the past year.

We've changed the clock setup to both the old a new (as we use the same binary) to 528MHz.

Now the HyperRAM isn't working at even 200MHz on old (600MHz, now clocked at 528) part. Core locks up as soon as the hyper RAM is accessed. We've had to back it off to 182MHz.

Is there some weird clock phasing issue between the core and FlexSPI peripheral?

To get 528MHz the ARM PLL was changed from *50 to *44.

Here is the current (working) setting:

p_shep_0-1666190392757.png

Changing PLL2_PFD2 to /24 will cause the instability.

Any ideas as to what the problem my be?

[jay_heng]

It is interesting finding.

You use 'PLL2_PFD2 to /22' in 600MHz RT1064 variety, and 216MHz HyperRAM is working properly?

[p_shep]

Correct, that the way we've been using the Cypress part for the last year - so overclocked a little.

Now with the changed core clock (from 600->528), the HyperRAM doesn't even work at the rated 200MHz (well, 198MHz, as it works out).

Further curiosity is that after boot-looping for anywhere from a couple of times to 30 mins later, it'll eventually spring into life and work fine.

On our new devices, not only is the RT1064 changed, but so is the HyperRAM to a winbond part. Initially we thought it was less tolerant refusing to work at 216MHz. But after making the same software clock changes to the old devices, we found the same issues.

Very odd indeed.