LPC5516 crach when running faster than 100Mhz

[jeroenbussink]

We have 2 batches of PCB's with different date codes on the CPU the one with sSD21480A  (week 48 2021) works without any problems the new batch has a date code sSD21490A  (week 49 2021) in this version the programs craches after the init.

To debug I have created a new default project with only a clock setup and this has the same problem it runs when I set the ferqeuency < 100Mhz and it crashes when its set > 100Mhz.

Any sugestions what can be done to solve this?

[frank_m]

I suspect you Flash settings are incorrect. If the number of wait cycles is too small, the core will hardfault.

The datasheet guarantees proper function at certain clock frequencies with a defined number of wait cycles, and at a given temperature. Each shipped part is guaranteed to meet this requirement, but many do exceed this guaranted values. How much depends on intrinsic parameters (batch), and on temperature.

[Pavel_Hernandez]

Hello,

I need more information about your MUCs, Could you give me a picture of both MCUs? Where did you buy the last lot?

Best regards,
Pavel

[jeroenbussink]

Hello,

I have attached a picture of a working and a not working CPU, both CPU are bought from AVNET.

[jsoccerd]

I know with the LPC5536 that the default flash memory speed is determined by the FMC register in SYSCON so if you don't change that it causes a hard fault on that processor. Check for similar registers and see if you can change the flash speed.

[jeroenbussink]

I have checked the FCM register but its set by the libary to the correct value (10)

[jsoccerd]

The value in the FLASHTIM should be 0xB or 11, not 10 if you want 150MHZ.

[jeroenbussink]

The system is running at 144Mhz, and then 10 should be ok according to the data sheet and the SDK functions. 

[jeroenbussink]

To test I increased the wait states to 11 and now the boards that are running at 144Mhz are working. Strange thing is that this value is other than calculated by the SDK and shown in the manual.

In a second test I put the sampels to a speed 150Mhz so the wait states are calculated at 11 now some of the boards still work and some of them stopped working again.

[jsoccerd]

That's probably because quartz oscillators have some inherent error in them, and the PLL also has some error, so I would suggest to go a little bit below a desired frequency to account for tolerances.

[jeroenbussink]

For now we have choosen a number of wait states one higher then stated in the manual

[frank_m]

If you think that products do no satisfy guaranteed requirements as described in the datasheet, you might contact the supplier (NXP) directly or via official distributor, not via a public forum.

I know of no vendor that discusses quality issues, unreleased products, or future portfolios on public fora.