About CC430 errata "RTC3" "Unreliabe write to RTC register"

Jianyi Bao said:

Only RTC in calendar mode needs  the workaround? what about counter mode?

errors occurs during access (writing) to RTC registers when RTC is in use

Jianyi Bao said:

The workaround only provides set functions like "SetRTCHOUR", "SetRTCMIN" and "SetRTCSEC", but no functions like "SetRTCTIM0", although they should write to the same registers, simply different for byte or word operation.

look in code:

;-------------------------------------------------------------------------------	
; int SetRTCMIN(int min)
; min will be set in the RTCMIN register
; Workaround requires to write 16 bits into RTCTIM0
; return value is read back from the RTCDOW register
; argument min is in R12
; return value is in R12
SetRTCMIN   push.w  R5                      ; push R5
            mov.b   &RTCSEC, R5             ; read RTCSEC into R5
            swpb    R12                     ; R12 holds min - swap lower and upper byte
            bis.w   R12, R5                 ; combine read RTCMIN and new RTCSEC
            dint
            nop
            mov.w   R5,&RTCTIM0             ; write to RTCTIM0
            nop
            eint
            mov.b   &RTCMIN, R12            ; Read RTCDOW register
            pop.w   R5                      ; pop R5
            reta
;-------------------------------------------------------------------------------	
; int SetRTCSEC(int sec)
; sec will be set in the RTCSEC register
; Workaround requires to write 16 bits into RTCTIM0
; return value is read back from the RTCSEC register
; argument sec is in R12
; return value is in R12
SetRTCSEC                                   ; R12 holds day in lower byte
            push.w  R5                      ; push R5
            mov.b   &RTCMIN, R5             ; read RTCMIN into lower byte of R5
            swpb    R5                      ; min is now in higher byte of R5
            bis.w   R12, R5                 ; combine read RTCMIN and new RTCSEC
            dint
            nop
            nop                             ; WG:
            mov.w   R5,&RTCTIM0             ; write to RTCTIM0		
            nop
            eint
            mov.b   &RTCSEC, R12            ; Read RTCSEC register
            pop.w   R5                      ; pop R5
            reta

Jianyi Bao said:

Even if it has provided read functions "GetRTCTIM0" and "GetRTCTIM1", in order to correctly read RTC count in counter mode, it still needs SW majority check to ensure the result is valid, for example by calling "GetRTCTIM0" twice and think the result is valid only when both returned value are the same?  Or, "GetRTCTIM0" will ensure the result will be valid?

"GetRTCTIM0" and "GetRTCTIM1"  functions are atomic they operating on 16 bit registers,

look in code, only word read instruction (mov.w) is used

; Read the RTC registers RTCTIM0
; return the values
GetRTCTIM0
            mov.w   &RTCTIM0,R12
            reta

; Read the RTC register RTCTIM1
; return the values
GetRTCTIM1
            mov.w   &RTCTIM1,R12
            reta

look in errata:

A write access to the RTC registers (SEC, MIN, HOUR, DATE, MON, YEAR, DOW) may
result in unexpected results.

so read access to the RTC should be ok,

1. For write RTC count, so you mean this workaround apply for both counter mode and calendar mode?

2. For read issue, please see the description in user's guide:

Accessing the RTCNT1, RTCNT2, RTCNT3, RTCNT4, RT0PS, RT1PS registers
When the counter clock is asynchronous to the CPU clock, any read from any RTCNT1,
RTCNT2, RTCNT3, RTCNT4, RT0PS, or RT1PS register should occur while the counter is
not operating. Otherwise, the results may be unpredictable. Alternatively, the counter may be
read multiple times while operating, and a majority vote taken in software to determine the
correct reading.

So I think the workaround didn't solve this issue, or maybe it can't be solved totally. Am I right?

yes you have right

when RTC clock is asynchronous to the CPU clock,
RTC value can change during RTC value is reading
and in this situation read value will be corrupted

like you wrote earlier, when you read value from RTC
you should make it two, or more times
and you should compare value from first read to value from next read, they should be equal

Thanks. Then I have some questions about the code. See the code and my questions below:

SetRTCMON:  push.w  R5                      ; push R5
            mov.b   &RTCDAY,R5              ; read RTCDAY into R5 ***Here only one read. So this means this function could be calle only when RTC is not running (by setting RTCHOLD?)***
            swpb    R12                     ; R12 holds month - swap lower and upper byte
            bis.w   R12,R5                  ; combine read RTCDAY and new RTCMON
            dint
            nop
            mov.w   R5,&RTCDATE             ; write to RTCDATE  
            nop
            nop
            nop
            nop
            eint
            mov.b   &RTCMON,R12             ; Read RTCMON register ***Why reading is needed here? Simply because we want to read the value back? That is, if we only need to set the value, can we remove this line the above nops when we want the code more compact?***
            pop.w   R5                      ; pop R5
            .if     $isdefed("__LARGE_CODE_MODEL__") = 0
            ret
            .else
            reta
            .endif