Errata Sheet Rel. 1.0, 09.02.2007 Device XC866L-2FR Marking/Step BE Package PG-TSSOP-38 This Errata Sheet describes the deviations from the current user documentation.The module oriented classification and numbering system uses an ascending sequence over several derivatives, including already solved deviations. So gaps inside this enumeration can occur. Table 1 Current Documentation XC866 User’s Manual V1.2 June 2006 XC866 Data Sheet V1.0 Feb 2006 Each erratum identifier follows the pattern Module_Arch.TypeNumber: • • • • Module: subsystem or peripheral affected by the erratum Arch: microcontroller architecture where the erratum was firstly detected. – AI: Architecture Independent (detected on module level) – CIC: Companion ICs – TC: TriCore (32 bit) – X: XC1xx / XC2xx (16 bit) – XC8: XC8xx (8 bit) – none: C16x (16 bit) Type: none - Functional Deviation; 'P' - Parametric Deviation; 'H' Application Hint; 'D' - Documentation Update Number: ascending sequencial number within the three previous fields. As this sequence is used over several derivatives, including already solved deviations, gaps inside this enumeration can occur. XC866L-2FR, BE 1/24 Rel. 1.0, 09.02.2007 Errata Sheet Note: Devices marked with EES or ES are engineering samples which may not be completely tested in all functional and electrical characteristics, therefore they should be used for evaluation only. The specific test conditions for EES and ES are documented in a separate Status Sheet. XC866L-2FR, BE 2/24 Rel. 1.0, 09.02.2007 Errata Sheet History List / Change Summary 1 History List / Change Summary Table 2 History List Version Date Table 3 Errata fixed in this step Remark Errata Short Description Chg EVR_XC8.005 Reset toggling issue for repeated power up Fixed EVR_XC8.H001 Consideration for Application with Fast Toggling Pin(s) Fixed Table 4 Functional Deviations Functional Deviation Short Description BRG_XC8.001 FDRES register is reloaded incorrectly 6 BROM_XC8.006 IRAM data is corrupted after any warm reset 6 BROM_XC8.010 SYSCON0.RMAP Switching Error 7 INT_XC8.003 Unintended External Interrupt 0/1 Request in Bypass Edge Detection Mode 8 INT_XC8.004 Unable to Detect New Interrupt Request if Any One of Timer 2 Interrupt Flags Is Not Cleared (Unexpectedly) 8 INT_XC8.005 Write to IRCON0 Blocks Interrupt Request of External Interrupt 0,1 11 XC866L-2FR, BE 3/24 Chg Pg Rel. 1.0, 09.02.2007 Errata Sheet History List / Change Summary Table 4 Functional Deviations Functional Deviation Short Description INT_XC8.006 Write to IRCON1 Blocks Concurrent Hardware Set of EIR Flag 11 OCDS_XC8.002 Protect Mask Not Implemented 12 OCDS_XC8.006 Last Break Address May Cause Program Flash to Fail 12 OCDS_XC8.007 Watchdog Timer behavior during Debug 13 OCDS_XC8.009 Break while CPU is in NMI service mode 13 OSC_XC8.002 OSC_CON register is not reset following a Upd ate wake-up reset 14 PLL_XC8.001 PLL N-Divider is reset to default value after Upd ate a WDT reset 14 WDT_XC8.002 WDT should not be refreshed during Prewarning period Upd ate 15 WDT_XC8.003 WDTRST bit cannot be set if software writes to PMCON0 register at the same time Upd ate 15 Table 5 Chg Pg Deviations from Electrical- and Timing Specification AC/DC/ADC Deviation Short Description AC_XC8.001 Oscillator long term frequency deviation ESD_XC8.001 ESD Human Body Model robustness XC866L-2FR, BE 4/24 Chg Pg 16 Upd ate 16 Rel. 1.0, 09.02.2007 Errata Sheet History List / Change Summary Table 6 Application Hints Hint Short Description ADC_XC8.H001 Arbitration mode when using external trigger at the selected input line REQTR 18 BROM_XC8.H001 SYSCON0.RMAP handling in ISR 18 DC_XC8.H001 VDDP 3.3V and 5V Applications 19 INT_XC8.H001 Interrupt Request With No Interrupt Flag Set 19 INT_XC8.H002 Effect of Interrupt Node Enable Bit on Interrupt Behavior 20 INT_XC8.H003 Interrupt Flags of External Interrupt 0 and 1 21 LIN_XC8.H001 LIN BRK field detection logic 22 LIN_XC8.H002 EOFSYN and ERRSYN bits status after exit from Boot ROM 23 OCDS_XC8.H002 Any NMI request is lost on Debug entry and during Debug 24 XC866L-2FR, BE 5/24 Chg Pg Rel. 1.0, 09.02.2007 Errata Sheet Functional Deviations 2 Functional Deviations BRG_XC8.001 FDRES register is reloaded incorrectly The reload of FDRES register happens when: overflow occurs or FDCON.FDEN bit is changed from 0 to 1 However, an unintended reload can also occur while FDEN bit is set. This happens whenever a write is carried out on FDCON register and as a result, FDEN bit is written again with 1, which reloads FDRES register. In Normal Divider Mode, this can be caused by the clearing of the overflow flag (FDCON.NDOV) by software, which indirectly sets the FDEN bit. The user should not operate the BRG in Normal Divider Mode. If a timer function is required, the user is recommended to use any of the available Timer modules. In Fractional Divider Mode, user should only write to FDCON when BRG is not running. If software writes to FDCON register while receiving a byte, the same unintended reload happens. However, the maximum error possible in this case is only 1/fMOD per write access to FDCON register and therefore, should not lead to a wrong reception. Workaround None. BROM_XC8.006 IRAM data is corrupted after any warm reset After any warm reset (i.e. reset without powering off the device), boot up via User Mode affects certain IRAM data. XC866L-2FR, BE 6/24 Rel. 1.0, 09.02.2007 Errata Sheet Functional Deviations The affected IRAM address ranges are: (1) 00H - 09H (2) 3EH - 40H (3) 61H - 68H (4) 80H - 0BFH Workaround None BROM_XC8.010 SYSCON0.RMAP Switching Error When executing from XRAM, if SYSCON0.RMAP is switched using an onemachine-cycle read-modify-write instruction (e.g. ORL dir,A) and the SFR is accessed immediately by an one-machine-cycle instruction (e.g. MOV A,dir) or a PUSH instruction, the SFR from the previous mapping might be accessed instead. This RMAP switching error does not occur if code is executed from the Flash memory. Workaround When executing code from XRAM, use two-machine-cycle instructions to either switch RMAP or access the SFR. Alternatively, add one or more instructions (e.g. NOP) between the one-machine-cycle RMAP switching and SFR accessing instructions. XC866L-2FR, BE 7/24 Rel. 1.0, 09.02.2007 Errata Sheet Functional Deviations INT_XC8.003 Unintended External Interrupt 0/1 Request in Bypass Edge Detection Mode For external interrupt 0 or 1 in bypass edge detection mode with falling edge as the active event, clearing the IRCON0 EXINT0/1 flag while the pin input is still low will lead to the internal circuitry falsely detecting a falling edge event. This will lead to unintended interrupt request. Workaround Never clear the IRCON0.EXINT0/1 flag in bypass edge detection mode. Access the TCON.IE0/1 flag directly instead, if necessary (TCON.IE0/1 flag is cleared automatically by hardware on vectoring to the interrupt routine). INT_XC8.004 Unable to Detect New Interrupt Request if Any One of Timer 2 Interrupt Flags Is Not Cleared (Unexpectedly) As illustrated in the simplified figure, the Timer 2 interrupt flags TF2 and EXF2 are combined as one interrupt request output. These flags are located within the Timer 2 kernel, with a single interrupt request line as output from the kernel. Timer 2 Overflow TF2 >=1 T2EX interrupt detection EXF2 EXEN2 EDGES EL Figure 1 Being of interrupt structure 2, the interrupt request of Timer 2 is detected on the rising edge of a positive pulse. XC866L-2FR, BE 8/24 Rel. 1.0, 09.02.2007 Errata Sheet Functional Deviations The problem is that it may occur at some point in the application, that any new Timer 2 interrupt request can no longer be detected after a return-from-interrupt (reti) due to service of earlier event(s). This happens when the following conditions are true: 1. Timer 2 events are serviced by interrupt (i.e. flags are checked and cleared only in the interrupt routine ISR), 2. Either TF2 or EXF2 is set at any one time throughout the ISR (even while the other is cleared) such that at least one of the flags is still set after reti, causing the Timer 2 request line, which is an OR-function of the two flags TF2 and EXF2, to remain set throughout the ISR and after reti. Two example scenarios are illustrated in the following figure: TF2 OR T2 R e q u e st EXF2 TF2 E .g . 1 EXF2 T 2 re q u e st IS R e n te re d IS R e xite d TF2 E .g . 2 EXF2 T 2 re q u e st Figure 2 This means, any future Timer 2 TF2 or EXF2 event is not able to cause a rising edge and therefore not able to trigger an interrupt request to the core – as if the Timer 2 interrupts had been disabled, until both EXF2 and TF2 flags are cleared at some time. The clearing of flags would have to be done by user’s code additionally outside of the Timer 2 interrupt routine, which is however normally not feasible with an interrupt service scheme. XC866L-2FR, BE 9/24 Rel. 1.0, 09.02.2007 Errata Sheet Functional Deviations Note: This condition affects only the detection of Timer 2 interrupt events TF2 and EXF2. It does not block the detection of other interrupt events belonging to the same interrupt node. Workaround There is no problem if EXF2 is not enabled by setting bit EXEN2 = 0 i.e. external events are disabled, or if bit DCEN = 1 in auto-reload mode. Otherwise if EXF2 is enabled, while TF2 is always enabled when Timer 2 is running, there are two suggested workaround: 1. If other events of Timer 2 interrupt node (ET2) need not be enabled for interrupt, disable this interrupt node and use software polling of the flags instead. 2. Before return from interrupt, check again if TF2 or EXF2 is (still) set (due to new request since the last check). If so, jump and execute the ISR routine from start. Exit only when all flags are checked to be cleared. However, dummy interrupt of the node may occur after return from interrupt, and should be ignored. Another drawback is if Timer 2 events are occurring at high rate, the CPU may be ’stuck’ in the service routine of the Timer 2 interrupt for a long time. <entry point of interrupt node service routine> ….. Start: check flag TF2 ….. clear flag TF2 ….. check flag EXF2 ….. clear flag EXF2 ….. Finish: if TF2 or EXF2 is set, jum p to Start reti (return from interrupt) Figure 3 XC866L-2FR, BE 10/24 Rel. 1.0, 09.02.2007 Errata Sheet Functional Deviations INT_XC8.005 Write to IRCON0 Blocks Interrupt Request of External Interrupt 0,1 Any write (read-modify-write or direct MOV) to the SFR IRCON0 will block an incoming interrupt request from external interrupt 0 or 1, even though the respective flag (EXINT0 or EXINT1) is set. Workaround After any write to the IRCON0, check (read IRCON0) the flags EXINT0 and EXINT1. If any flag is set, run the service routine; otherwise proceed. In case of enabled for interrupt, the service routine should be duplicated: one copy as the interrupt service routine (with reti executed); another copy in main code memory for software call (with ret executed). INT_XC8.006 Write to IRCON1 Blocks Concurrent Hardware Set of EIR Flag Any write (read-modify-write or direct MOV) to the SFR IRCON1 will block a concurrent hardware set of the SSC error interrupt flag EIR (due to event happening). On the other hand, the interrupt request is triggered by the EIR event nevertheless. Workaround In the interrupt service routine, check and clear the kernel error source flags (TE, RE, PE and BE) instead of the EIR flag. XC866L-2FR, BE 11/24 Rel. 1.0, 09.02.2007 Errata Sheet Functional Deviations OCDS_XC8.002 Protect Mask Not Implemented The protect mask feature is not implemented for the OCDS SFRs. This means it is possible for rwh SFR bits to not hold an updated (in a way, become corrupted) value under the following circumstances of an application: The hardware updates any rwh bit of a SFR which is in the midst of a read-modifywrite operation. The consequence is that the SFR bit update by hardware is lost, as the bit value is overwritten by the write-back operation. Workaround Software to take care to not include read-modify-write operations to OCDS SFRs which contain rwh bits. Use alternative instruction(s) instead. OCDS_XC8.006 Last Break Address May Cause Program Flash to Fail This problem occurs only if the last break address results in a return address in the range 0000H to 00FFH. In this case, the Flash bank where Flash wordline program is executed will hang in the program state while no Flash NMI is triggered. The contents of the Flash wordline may also be corrupted. This renders the particular Flash bank unusable for code fetch or data read. Workaround When debugging code that calls Flash programming, do not break the CPU that results in the return address (from debug mode) to be in the range 0000H to 00FFH. Note: The return address is a function of the break address, refer to OCDS chapter of User’s Manual. XC866L-2FR, BE 12/24 Rel. 1.0, 09.02.2007 Errata Sheet Functional Deviations OCDS_XC8.007 Watchdog Timer behavior during Debug Firstly, the Watchdog Timer (WDT) cannot be refreshed by writing WDTCON.WDTRS in debug mode. Secondly, on entering debug mode, the timer base counter which provide the clock to WDT is cleared to zero. The effect is that on exiting Monitor Mode in user mode, the WDT will count a little longer before overflow. This extra time is less than one WDT count. Note: In debug mode, the WDT is suspended and stops counting. Workaround None. These WDT behavior occur only during debug. OCDS_XC8.009 Break while CPU is in NMI service mode While the CPU is in NMI service mode, the debug function has the following limitations: 1. External break via JTAG or MBC pin occurring while in NMI service mode, results in debug mode entry being postponed until NMI mode is ended. 2. Hardware or software breakpoints inside the NMI service routine will trigger the debug function. However, it is not possible to return to and execute user code (run or step) at the point of break as the debug cannot be exit. Workaround The workaround for the respective scenerios are 1. None. In fact, since NMI is usually a critical situation, any asynchronous break event postponed till after return from NMI mode has insignificant effects. 2. Do not set breakpoints or put the TRAP instruction inside the NMI service routine. XC866L-2FR, BE 13/24 Rel. 1.0, 09.02.2007 Errata Sheet Functional Deviations OSC_XC8.002 OSC_CON register is not reset following a wake-up reset On a wake-up from power-down with reset, the register OSC_CON is not reset to its default value as specified. If an external oscillator is used, it will remain as the selected oscillator source on wake-up reset. On the other hand, the PLL settings including the NDIV value will be reset to the default state. Depending on the external oscillator’s frequency (fOSC), there are three possible scenarios: 1. If 9.375 MHz ≤ fOSC ≤ 12 MHz, the system starts executing the user code correctly, even though it might run at a reduced frequency. 2. If fOSC < 9.375 MHz, the system might hang in an endless loop due to the inability of the PLL to lock. Workaround The wake-up from power-down with reset feature should not be used with an external oscillator outside the range of 9.375 to 12 MHz. There is no impact on the user if the system is using the on-chip oscillator. PLL_XC8.001 PLL N-Divider is reset to default value after a WDT reset The user configured PLL N-divider value should be retained following a WDT reset. However in the current implementation, it is reset to its default value instead. A problem arises when an external oscillator is used to generate the system clock. Depending on the external oscillator’s frequency (fOSC), there are three possible scenarios: 1. If 9.375 MHz ≤ fOSC ≤ 12 MHz, the system starts executing the user code correctly, even though it might run at a reduced frequency. 2. If fOSC < 9.375 MHz, the system might hang in an endless loop due to the inability of the PLL to lock. Workaround The WDT should not be used with an external oscillator outside the range of 9.375 to 12 MHz (scenario 2). If the external oscillator falls within 9.375 to 12 XC866L-2FR, BE 14/24 Rel. 1.0, 09.02.2007 Errata Sheet Functional Deviations MHz (scenario 1), the PLL N-divider must always be re-initialized in the user code with the required value after a WDT reset. There is no impact on the user if the system is using the on-chip oscillator. WDT_XC8.002 WDT should not be refreshed during Prewarning period If the WDT is not serviced before the timer overflows, a WDT NMI request is asserted and Prewarning is entered. During the Prewarning period, which last for 30H, refreshing of the Watchdog Timer should not be possible and the microcontroller is expected to reset at the end of this period. However in the current implementation, the WDT reset can be prevented by constantly writing to WDTCON.WDTRS during Prewarning. Workaround The user should not refresh the WDT upon entering the Prewarning period. WDT_XC8.003 WDTRST bit cannot be set if software writes to PMCON0 register at the same time When a software write access to register PMCON0 happens at the same time as a WDT reset, the WDT reset indication bit, PMCON0.WDTRST, may not be updated correctly. Workaround The WDT Prewarning feature should always be enabled (bit NMICON.NMIWDT is set to enable WDT NMI) so that the WDT NMI routine is always entered before a WDT reset. In the WDT NMI routine, care must be taken not to write to the PMCON0 register. XC866L-2FR, BE 15/24 Rel. 1.0, 09.02.2007 Errata Sheet Deviations from Electrical- and Timing Specification 3 Deviations from Electrical- and Timing Specification AC_XC8.001 Oscillator long term frequency deviation Table 7 Parameter Symbol Limit Values Unit Test Condition min. typ. max. Long term frequency ∆fLT deviation1) -5.0 - 5.0 % with respect to fNOM+ ∆fCC, over lifetime and temperature (-10°C to 125°C), for one device after trimming. -6.0 - 0 % with respect to fNOM+∆fCC, over lifetime and temperature (-40°C to -10°C), for one device after trimming. 1) Not all parameters are 100% tested, but are verified by design/characterisation and test correlation. Workaround None. ESD_XC8.001 ESD Human Body Model robustness ESD susceptibility according to Human Body Model (HBM) is up to 600V for VDDC pin; and up to 2000V for all other pins. XC866L-2FR, BE 16/24 Rel. 1.0, 09.02.2007 Errata Sheet Deviations from Electrical- and Timing Specification Workaround None. XC866L-2FR, BE 17/24 Rel. 1.0, 09.02.2007 Errata Sheet Application Hints 4 Application Hints ADC_XC8.H001 Arbitration mode when using external trigger at the selected input line REQTR If an external trigger is expected at the selected input line REQTR to trigger a pending request, the arbitration mode should be set (PRAR.ARBM=1) where the arbitration is started by pending conversion request. This selection will minimize the jitter between asynchronous external trigger with respect to the arbiter and the start of the conversion. The jitter can only be minimized while no other conversion is running and no higher priority conversion can cancel the triggered conversion. In this case, a constant delay (no jitter) has to be taken into account between the trigger event and the start of the conversion. BROM_XC8.H001 SYSCON0.RMAP handling in ISR The ISR has to handle SYSCON0.RMAP correctly when Flash user routines provided in the Boot ROM are used together with the interrupt system. Any ISR with the possibility of interrupting these user routines has to do the following in the interrupt routine: save the value of the RMAP bit at the beginning restore the value before the exit This is to prevent access of the wrong address map upon return to the Flash user routine since the RMAP bit may be changed within the interrupt routine. The critical point is when Flash user routines sets RMAP to ‘1’ and the interrupt occurs that needs RMAP at ‘0’ in the ISR. Please note that NMI is an interrupt as well. XC866L-2FR, BE 18/24 Rel. 1.0, 09.02.2007 Errata Sheet Application Hints DC_XC8.H001 VDDP 3.3V and 5V Applications For XC866-2FR and XC866-4FR flash devices, two marking options VDDP 3.3V and VDDP 5V are provided. The correct option must be chosen according to the voltage supply of the application. INT_XC8.H001 Interrupt Request With No Interrupt Flag Set It might occur in the application, that sometimes interrupt requests are serviced, but no active interrupt flags are found. This would happen for those interrupt nodes which is shared by several interrupt sources. Consider the following interrupt service routine pseudo-code, and scenerio where a interrupt source A event leads to interrupt request of the node and CPU vectors to the interrupt routine. Meanwhile, interrupt source B and its flag becomes active any time in the duration indicated by T: <entry point of interrupt node service routine> ….. check flag A T ….. clear flag A ….. check flag B ….. clear flag B ….. reti (return from interrupt) Figure 4 In this case, flag B will be cleared as a standard procedure by the interrupt routine in the current service of the interrupt node. However, the pending interrupt request for the node remains activated after RETI, as it is only cleared XC866L-2FR, BE 19/24 Rel. 1.0, 09.02.2007 Errata Sheet Application Hints by hardware when CPU acknowledge the interrupt and vectors to the interrupt routine. This leads to following servicing of the interrupt node again, but potentially no active flag is found, i.e. dummy interrupt service. The point to note is that the interrupt source B is not lost, as it was actually serviced in the current service of the interrupt node. The recommendation is to ignore these dummy interrupt vectoring. INT_XC8.H002 Effect of Interrupt Node Enable Bit on Interrupt Behavior A-Step When the interrupt event occurs while the respective interrupt node enable bit (IENx.y) is disabled, the interrupt status flag will be set and the interrupt request will be captured. Hence, enabling the respective IENx.y bit later will result in the CPU servicing of the interrupt node. It is not possible to clear any pending interrupt request by software. Refer to INT_XC8.001 on the Dummy Interrupt issue for further detail. B-Step When the interrupt event occurs while the respective IENx.y bit is disabled, the interrupt status flag will be set but the interrupt request will not be captured. Hence, enabling the respective IENx.y bit later will not result in the CPU servicing the interrupt node for the prior interrupt event. Clearing the IENx.y bit of a node will clear its pending interrupt request. In both steps, the interrupt enable bit (EA) may be used to globally disable all interrupts. All incoming interrupt requests will be captured in respective interrupt nodes provided the IENx.y bit is enabled. All interrupt requests will be serviced according to priority when EA bit is enabled. XC866L-2FR, BE 20/24 Rel. 1.0, 09.02.2007 Errata Sheet Application Hints INT_XC8.H003 Interrupt Flags of External Interrupt 0 and 1 External interrupt 0 and 1 may individually be selected via respective bits (EXINTx) in EXICON0 register, to request interrupt on falling edge, rising edge, both edges or to bypass the edge detection. EINT0 EXINT0 IE0 IRCON0.0 TCON.1 IT0 0003 H IEN0.0 TCON.0 EXINT0 EX0 to CPU EXICON0.0/1 EINT1 EXINT1 IE1 IRCON0.1 TCON.3 IT1 EXINT1 TCON.2 EXICON0.2/3 EX1 0013 H IEN0.2 EA IEN0.7 Figure 5 Edge detection is done in the system unit. If enabled, an active event will set the EXINTx flag and correspondingly set the IEx flag in TCON. It should be noted that after any external interrupt x event, flag EXINTx must be cleared. In case of falling edge as active event, this allows any future active event to be able to set the flag IEx as interrupt request. In case of low level as active event, this prevents unintended recurring triggering of interrupt request. In case of bypass edge detection, the input is connected directly to the core (bypass the system unit). The flag IEx in TCON will be set on the selected falling edge or low level (at least 2 clock cycles) event. In case of falling edge selected as active event by TCON.ITx, the flag EXINTx is set in addition to the flag IEx. However to use this mode properly, user must not clear the flag EXINTx and if necessary, to only access the flag TCON.IEx (refer to INT_XC8.003). XC866L-2FR, BE 21/24 Rel. 1.0, 09.02.2007 Errata Sheet Application Hints Besides the above notes, the following should be noted on the behavior regarding setting and clearing of the external interrupt x (x = 0 or 1) flags, applicable to both edge and bypass edge detection modes: Setting of External Interrupt x Flags 1. The flag TCON.IEx will be set in all modes selectable via EXICON0 register. 2. Flag IRCON0.EXINTx will be set in all modes as long as an active edge is detected; flag EXINTx will not be set for low level as active event. Clearing of External Interrupt x Flags 1. Flag IEx is cleared automatically by hardware when the interrupt is being vectored to. 2. Flag EXINTx has to be cleared by software. 3. Clearing one external interrupt x flag will not clear the other. Especially, clearing flag EXINTx will not clear the flag IEx. Being of interrupt structure 1, the flag IEx is the request polled by the CPU for interrupt servicing. Therefore user has to take care to clear the flag IEx before switching from SW polling method to enabling the external interrupt x node, to prevent potential dummy interrupt request. 4. Always clear both EXINTx and IEx flags before (if) changing the trigger select in EXICON0 register. LIN_XC8.H001 LIN BRK field detection logic Based on the hardware implementation, the maximum number of bits in the BRK field must follow the formula: Maximum number of bits in BRK field = Baud Rate x 4095 Sample Frequency PCLK Sample Frequency = 8x 2 XC866L-2FR, BE 22/24 BGSEL Rel. 1.0, 09.02.2007 Errata Sheet Application Hints For example, if LIN baudrate is 19.2kbps, BGSEL = 0 and CPU frequency is 26.67MHz, the maximum number of bits in BRK field would be: 19.2k x 4095 / ( (26.67M / 8) ) = ~23.6 bits If the maximum number of bits in the BRK field exceeded, the internal counter will overflow which results in baudrate detection error. Therefore, the user is advised to choose the appropriate BGSEL value for the required baudrate detection range. The calculated value above does not consider sample error and transmission error, nevertheless it can be used as a guideline. LIN_XC8.H002 EOFSYN and ERRSYN bits status after exit from Boot ROM When the device exits from Boot ROM and starts user code execution (either in XRAM or flash memory), the bitfields EOFSYN and ERRSYN in SFR FDCON remain in their last accessed states. Therefore these bitfields must be initialized according to the application requirement by the user code. The following are the possible application scenarios that could lead to the above: 1) When any LIN BSL supported modes is activated before the device starts to execute user code (either in XRAM or flash memory). 2) After the device is reset and remain in the LIN BSL, if there is any error frame detected in the LIN bus, the user code in the flash memory will be executed. User code should initialize the register properly according to the application requirement. XC866L-2FR, BE 23/24 Rel. 1.0, 09.02.2007 Errata Sheet Application Hints OCDS_XC8.H002 Any NMI request is lost on Debug entry and during Debug All NMI events are disabled while in debug mode. This has two main effects: 1. On debug entry, any pending NMI request will be lost, although the status flag remains set. The probability of losing an NMI request in this way is very low, since NMI always has the highest priority to be serviced. 2. Any NMI event that occurs during debugging is not able to generate an NMI request (event interrupt is lost) although the status flag will be set. It is normally not critical that on exit from debug mode, the CPU must service NMI requests that had occurred while in debug mode. The fact that the debug system is not specified to support NMI interrupt while in debug mode makes the above trivial. As precaution, avoid starting any debug session while expecting an NMI event. XC866L-2FR, BE 24/24 Rel. 1.0, 09.02.2007