LPC4370 ADCHS high interrupt jitter

[vitaliylivnov]

Hello.

Our application uses HSADC and DMA to write samples to RAM. We also use threshold crossing interrupt to search signal edges.

We achieved the minimum HSADC ISR calling delay by moving the interrupt table and handler to RAM. Now it is about 100 ns. However, this value can only be obtained if the microcontroller is in interrupt wait mode (__WFI()) before it is triggered. If the microcontroller performs any functions before the interrupt is triggered, the delay time becomes unstable and can take values from 100 ns to 2 μs.

Using code optimization and transferring all software to RAM improves stability a little, but does not completely solve the problem. Tell us, please, how to get around this problem?

[xiangjun_rong]

Hi, Vitaliy,

You are right, for the MPU, if you configure region incorrectly, when you access the region, the MemManage exception will be triggered. If you don’t have an exception; accordingly, the region has the correct configuration.

BR

XiangJun Rong

[vitaliylivnov]

Hi. xiangjun.rong‌

That is, at the moment, the cause of the problem is unclear?
I noticed that MPU allows you to disable caching for different regions. Perhaps this can speed up the code and reduce jitter time of ISR call?

[xiangjun_rong]

Hi, Vitaliy,

I think your problem becomes the question how the LPC4370 can run in parallel, for example when the ADC interrupt happens, all the processes occur simultaneously. The DMA transfer data from ADC result register to memory, while the core fetches code from spifi flash or onm-chip RAM, the core writes registers to stack, so it is difficult.

Anyway, pls refer to the Fig 2. LPC4370 Block diagram (flashless parts, triple-core, 12-bit ADCHS) in UM10503.pdf, put the code into on-chip RAM and run in on-chip, while keep in mind that you have to use different bus so that all the process can run in parallel.

I do not know if it can help you.

BR

XiangJun Rong

[vitaliylivnov]

Hi, xiangjun.rong‌

Thanks for your answer. We tried to transfer all the code to RAM, and with this we turned off the parallel operation of DMA. Problems begin to occur as soon as the CPU performs any functions during the occurrence of an interrupt. And the more complex and voluminous the function, the greater the delay of the interrupt handler.

We thought that perhaps the CPU is caching the current instructions, and this takes time.

[xiangjun_rong]

Hi, Vitaliy,

As you know that the LPC4370 has MPU(Memory Protection Unit), you have to configure the SRAM as a region so that the SRAM location is code-fetch enabled.

Pls refer to the following section.

I copy the UM of Cortex-M4, which includes the MPU chapter.

Hope it can help you

BR

Xiangjun Rong

3.3.6 Memory Protection Unit (MPU)
The MPU is a integral part of the ARM Cortex-M4 for memory protection and supported by
all LPC43xx parts. The processor supports the standard ARMv7 Protected Memory
System Architecture model. The MPU provides full support for:
• protection regions
• overlapping protection regions, with ascending region priority (7 = highest priority, 0 =
lowest priority)
• access permissions
• exporting memory attributes to the system
MPU mismatches and permission violations invoke the programmable-priority
MemManage fault handler. See the ARMv7-M Architecture Reference Manual for more
information.
The access permission bits, TEX, C, B, AP, and XN, of the Region Access Control
Register control access to the corresponding memory region. If an access is made to an
area of memory without the required permissions, a permission fault is raised. For more
information, see the ARMv7-M Architecture Reference Manual.
The MPU is used to enforce privilege rules, to separate processes, and to enforce access
rules. For details on how to use the MPU and for the register description refer to the ARM
Cortex-M4 Technical Reference Manual.

[vitaliylivnov]

Hi, xiangjun.rong‌

Thanks for your answer. 

We have the following configuration of regions.

The code is executed from the RamLoc128 region. If I correctly understood the purpose of the MPU, it throws an exception if we try to execute code from a region that is not intended for this. We don’t have an exception; accordingly, our region has the correct configuration, right? Please correct me if I misunderstood you.