dsPIC33F RTSP问题

我使用的是在50 MHz下运行的DSPIC33 FJ64 GS606，一个基于Microchip的引导加载程序示例的引导加载程序。该过程涉及擦除应用程序存储器空间，然后加载一行程序数据（64个指令，192个字节），并使用RTSP程序行命令将数据写入闪存。使用16位CRC在输入上验证行数据，如果正确，行被编程。编程继续，直到所有行被加载和编程。引导加载程序将使用另一个16位CRC检查应用程序区域的完整性，如果正确，则跳转到应用程序。引导加载程序偶尔会失败，并且在检查之后，我会发现通常一行，有时几行没有编程。它们仍然处于擦除状态，读取为0x00 FFFFFF。我的应用程序占用约130行。在500个程序行尝试中，失败率大约为1。我使用以下的汇编程序来擦除页和程序行函数：这部分的单元耐久性至少是10000个擦除/写入周期，并且我还没有接近超过这个值。我已经检查了这部分的勘误表，没有提到RTSP问题。这个问题在几个不同的程序集上被观察到，几乎相同的失败率。在尝试的程序行事件之后，我将添加一个检查函数来尝试捕获这个失败模式。有人看到过这种行为，或者对我有什么建议？谢谢，文斯



      以下为原文

    I am using the dsPIC33FJ64GS606 running at 50 MHz, with a boot loader based on one of the boot loader examples from Microchip. The process involves erasing the application memory space, then loading one row of program data (64 instructions, 192 bytes) and using the RTSP Program Row command to write the data to the Flash. Row data is verified on input using a 16 bit CRC, and if correct the row is programmed. The programming continues until all rows are loaded and programmed. The boot loader will then check the integrity of the application area using another 16 bit CRC, and jump to the application if correct.
The boot loader will occasionally fail, and upon examination I will find that usually one, and sometimes several, rows did not program. They will still be in the erased state, read as 0x00FFFFFF.
My application program occupies about 130 rows. Failure rate is about 1 failed row in 500 Program Row attempts.
I am using the following assembly routine for the Erase Page and Program Row functions:
_FlashCmd:
mov W0,NVMCON
push SR ;save int status and disable
mov 0x00e0,W0
ior SR
mov #0x55,W0 ;Unlock sequence - interrupts need to be off
mov W0,NVMKEY
mov #0xAA,W0
mov W0,NVMKEY
bset NVMCON,#WR
nop ;Required
nop
flashcmdwait:
btsc NVMCON,#WR ;Wait for write end
bra flashcmdwait

Cell endurance for this part is a minimum of 10,000 Erase/Write cycles, and I have not come close to exceeding this value. I have checked the errata for this part, and see not mention of RTSP issues. This problem is observed on several different assemblies, at approximately the same failure rate.
I will be adding a checking function after the attempted Program Row event to try to capture this failure mode.
Has anyone seen this behavior, or have any suggestions for me?
Thanks,
Vince