C167CR-4R ES-DA, DA (historic/no update)

Microcontroller Components
Errata Sheet
November 23, 1998 / Release 1.2
Device:
SAK-C167CR-4RM
Stepping Code / Marking:
ES-DA, DA
Package:
MQFP-144
This Errata Sheet describes the deviations from the current user documentation. The
classification and numbering system is module oriented in a continual ascending sequence
over several derivatives, as well already solved deviations are included. So gaps inside this
enumeration could occur.
The current documentation is: Data Sheet: C167CR-4RM Data Sheet 07.97,
User’s Manual: C167 Derivatives User’s Manual V2.0 03.96
Instruction Set Manual 12.97 Version 1.2
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.
Change summary to Errata Sheet Rel.1.1 for devices with stepping code/marking
ES-DA, DA:
•
•
•
•
•
•
•
•
•
•
ADC Overload Current (ADCC.2.2): description modified (step specific)
Spikes on CS# lines after access with RDCS# and/or WRCS# (BUS.17)
PEC Transfers after JMPR (BUS.18)
Note about workaround for Tasking compiler added for problem CPU.16
Note on Interrupt Register Behaviour of CAN module added
Modifications of ADM field while bit ADST = 0 (ADC.11): reference to single channel
continuous mode in condition (1) eliminated
Read Access to XPERs in Visible Mode (X9): description modified
new naming convention for DC/AC specification deviations used
Output low voltage VOL (bus pins) tested according to specification (DC.VOL.1)
Input hysteresis HYS (special threshold) tested according to specification (DC.HYS.1)
Semiconductor Group
Errata Sheet, C167CR-4RM, ES-DA, DA, 1.2, Mh
- 1 of 15-
Functional Problems:
PWRDN.1:
Execution of PWRDN Instruction while pin NMI# = high
When instruction PWRDN is executed while pin NMI# is at a high level, power down mode should not
be entered, and the PWRDN instruction should be ignored. However, under the conditions described
below, the PWRDN instruction may not be ignored, and no further instructions are fetched from
external memory, i.e. the CPU is in a quasi-idle state. This problem will only occur in the following
situations:
a) the instructions following the PWRDN instruction are located in external memory, and a multiplexed
bus configuration with memory tristate waitstate (bit MTTCx = 0) is used, or
b) the instruction preceding the PWRDN instruction writes to external memory or an XPeripheral
(XRAM, CAN), and the instructions following the PWRDN instruction are located in external
memory. In this case, the problem will occur for any bus configuration.
Note: the on-chip peripherals are still working correctly, in particular the Watchdog Timer will reset the
device upon an overflow. Interrupts and PEC transfers, however, can not be processed. In case NMI#
is asserted low while the device is in this quasi-idle state, power down mode is entered.
Workaround:
Ensure that no instruction which writes to external memory or an XPeripheral precedes the PWRDN
instruction, otherwise insert e.g. a NOP instruction in front of PWRDN. When a muliplexed bus with
memory tristate waitstate is used, the PWRDN instruction should be executed out of internal RAM or
XRAM.
CPU.16:
Data read access with MOVB [Rn], mem instruction to internal
ROM/Flash/OTP
When the MOVB [Rn], mem instruction (opcode 0A4h) is executed, where
1. mem specifies a direct 16-bit byte operand address in the internal ROM/Flash memory,
AND
2. [Rn] points to an even byte address, while the contents of the word which includes the byte
addressed by mem is odd,
OR
[Rn] points to an odd byte address, while the contents of the word which includes the byte addressed
by mem is even
the following problem occurs:
a) when [Rn] points to external memory or to the X-Peripheral (XRAM, CAN, etc.) address space, the
data value which is written back is always 00h
b) when [Rn] points to the internal RAM or SFR/ESFR address space,
- the (correct) data value [mem] is written to [Rn]+1, i.e. to the odd byte address of the selected word in
case [Rn] points to an even byte address,
- the (correct) data value [mem] is written to [Rn]-1, i.e. to the even byte address of the selected word
in case [Rn] points to an odd byte address.
Workaround:
When mem is an address in internal ROM/Flash/OTP memory, substitute instruction
MOVB [Rn], mem
e.g. by
MOV Rm, #mem
MOVB [Rn], [Rm]
Semiconductor Group
Errata Sheet, C167CR-4RM, ES-DA, DA, 1.2, Mh
- 2 of 15-
Notes on compilers:
- the Keil C166 Compiler V3.10 has been extended by the directive FIXROM which avoids accesses
to ’const’ objects via the instruction MOVB [Rn], mem.
- the Tasking compiler provides a workaround for this problem from version V6.0r2 on
CPU.17:
Arithmetic Overflow by DIVLU instruction
For specific combinations of the values of the dividend (MDH,MDL) and divisor (Rn), the Overflow (V)
flag in the PSW may not be set for unsigned divide operations, although an overflow occurred.
E.g.:
MDH
F0F0
MDL
Rn
MDH MDL
0F0Fh : F0F0h = FFFF FFFFh, but no Overflow indicated!
(result with 32-bit precision: 1 0000h)
The same malfunction appears for the following combinations:
n0n0 0n0n : n0n0
n00n 0nn0 : n00n
n000 000n : n000
n0nn 0nnn : n0nn
where n means any Hex Digit between 8 ... F
i.e. all operand combinations where at least the most significant bit of the dividend (MDH) and the
divisor (Rn) is set.
In the cases where an overflow occurred after DIVLU, but the V flag is not set, the result in MDL is
equal to FFFFh.
Workaround:
Skip execution of DIVLU in case an overflow would occur, and explicitly set V = 1.
E.g.:
CMP Rn, MDH
JMPR cc_ugt, NoOverflow
; no overflow if Rn > MDH
BSET V
; set V = 1 if overflow would occur
JMPR cc_uc, NoDivide
; and skip DIVLU
NoOverflow:
DIVLU Rn
NoDivide:
...
; next instruction, may evaluate correct V flag
Note:
- the KEIL C compiler, run time libraries and operating system RTX166 do not generate or use
instruction sequences where the V flag in the PSW is tested after a DIVLU instruction.
- with the TASKING C166 compiler, for the following intrinsic functions code is generated which uses
the overflow flag for minimizing or maximizing the function result after a division with a DIVLU:
_div_u32u16_u16()
_div_s32u16_s16()
_div_s32u16_s32()
Consequently, an incorrect overflow flag (when clear instead of set) might affect the result of one of the
above intrinsic functions but only in a situation where no correct result could be calculated anyway.
These intrinsics first appeared in version 5.1r1 of the toolchain.
Libraries: not affected
Semiconductor Group
Errata Sheet, C167CR-4RM, ES-DA, DA, 1.2, Mh
- 3 of 15-
BUS.17:
Spikes on CS# Lines after access with RDCS# and/or WRCS#
Spikes of about 5 ns width (measured at VOH = 0.9 Vcc) from Vcc to Vss may occur on Port 6 lines
configured as CS# signals. The spikes occur on one CSx# line at a time for the first external bus
access which is performed via a specific BUSCONx/ADDRSELx register pair (x=1..4) or via BUSCON0
(x=0) when the following two conditions are met:
1. the previous bus cycle was performed in a non-multiplexed bus mode without tristate waitstate
via a different BUSCONy/ADDRSELy register pair (y=1..4, y≠x) or BUSCON0 (y=0, y≠x) and
2. the previous bus cycle was a read cycle with RDCSy# (bit BUSCONy.CSRENy = 1) or a write cycle
with WRCS# (bit BUSCONy.CSWENy = 1).
The position of the spikes is at the beginning of the new bus cycle which is performed via CSx#,
synchronous with the rising edge of ALE and synchronous with the rising edge of RD#/WR# of the
previous bus cycle.
Potential effects on applications:
-
when CS# lines are used as CE# signals for external memories, typically no problems are
expected, since the spikes occur after the rising edge of the RD# or WR# signal.
when CS# lines configured as RDCS# and/or WRCS# are used e.g. as OE# signals for external
devices or as clock input for shift registers, problems may occur (temporary bus contention for read
cycles, unexpected shift operations, etc.). When CS# lines configured as WRCS# are used as WE#
signals for external devices, no problems are expected, since a tristate waitstate should be used
anyway due to the negative address hold time after WRCS# (t55) without tristate WS.
Workarounds:
1. Use a memory tristate WS (i.e. leave bit BUSCONy.5 = 0) in all active BUSCON registers where
RD/WR-CS# is used (i.e. bit BUSCONy.CSRENy = 1 and/or bit BUSCONy.CSWENy = 1), or
2. Use Address-CS# instead of RD/WR-CS# (i.e. leave bits BUSCONy[15:14] = 00b) for all
BUSCONy registers where a non-multiplexed bus without tristate WS is configured (i.e. bit
BUSCONy.5 = 1).
BUS.18:
PEC Transfers after JMPR instruction
Problems may occur when a PEC transfer immediately follows a taken JMPR instruction when the
following sequence of 4 conditions is met (labels refer to following examples):
1. in an instruction sequence which represents a loop, a jump instruction (Label_B) which is capable
of loading the jump cache (JMPR, JMPA, JB/JNB/JBC/JNBS) is taken
2. the target of this jump instruction directly is a JMPR instruction (Label_C) which is also taken and
whose target is at address A (Label_A)
3. a PEC transfer occurs immediately after this JMPR instruction (Label_C)
4. in the following program flow, the JMPR instruction (Label_C) is taken a second time, and no other
JMPR, JMPA, JB/JNB/JBC/JNBS or instruction which has branched to a different code segment
(JMPS/CALLS) or interrupt has been processed in the meantime (i.e. the condition for a jump
cache hit for the JMPR instruction (Label_C) is true)
In this case, when the JMPR instruction (Label_C) is taken for the second time (as described in
condition 4 above), and the 2 words stored in the jump cache (word address A and A+2) have been
processed, the word at address A+2 is erroneously fetched and executed instead of the word at
address A+4.
Note: the problem does not occur when
- the jump instruction (Label_C) is a JMPA instruction
- the program sequence is executed from internal ROM/Flash
Semiconductor Group
Errata Sheet, C167CR-4RM, ES-DA, DA, 1.2, Mh
- 4 of 15-
Example1:
Label_A: instruction x
; Begin of Loop
instruction x+1
.....
Label_B: JMP Label_C ; JMP may be any of the following jump instructions:
JMPR cc_zz, JMPA cc_zz, JB/JNB/JBC/JNBS
; jump must be taken in loop iteration n
; jump must not be taken in loop iteration n+1
.....
Label_C: JMPR cc_xx, Label_A
; End of Loop
; instruction must be JMPR (single word instruction)
; jump must be taken in loop iteration n and n+1
; PEC transfer must occur in loop iteration n
Example2:
Label_A: instruction x
; Begin of Loop1
instruction x+1
.....
Label_C: JMPR cc_xx, Label_A
; End of Loop1, Begin of Loop2
; instruction must be JMPR (single word instruction)
; jump not taken in loop iteration n-1, i.e. Loop2 is entered
; jump must be taken in loop iteration n and n+1
; PEC transfer must occur in loop iteration n
.....
Label_B: JMP Label_C
; End of Loop2
; JMP may be any of the following jump instructions:
JMPR cc_zz, JMPA cc_zz, JB/JNB/JBC/JNBS
; jump taken in loop iteration n-1
A code sequence with the basic structure of Example1 was generated e.g. by a compiler for
comparison of double words (long variables).
Workarounds:
1. use a JMPA instruction instead of a JMPR instruction when this instruction can be the direct target
of a preceding JMPR, JMPA, JB/JNB/JBC/JNBS instruction, or
2. insert another instruction (e.g. NOP) as branch target when a JMPR instruction would be the direct
target of a preceding JMPR, JMPA, JB/JNB/JBC/JNBS instruction, or
3. change the loop structure such that instead of jumping from Label_B to Label_C and then to
Label_A, the jump from Label_B directly goes to Label_A.
Notes on compilers:
In the Hightec compiler beginning with version Gcc 2.7.2.1 for SAB C16x – V3.1 Rel. 1.1, patchlevel 5,
a switch –m bus18 is implemented as workaround for this problem. In addition, optimization has to be
set at least to level 1 with –u1.
The Keil C compiler and run time libraries do not generate or use instruction sequences where a JMPR
instruction can be the target of another jump instruction, i.e. the conditions for this problem do not
occur.
In the TASKING C166 Software Development Tools, the code sequence related to problem BUS.18
can be generated in Assembly. The problem can also be reproduced in C-language by using a
particular sequence of GOTOs.
With V6.0r3, TASKING tested all the Libraries, C-startup code and the extensive set of internal testsuite sources and the BUS.18 related code sequence appeared to be NOT GENERATED.
Semiconductor Group
Errata Sheet, C167CR-4RM, ES-DA, DA, 1.2, Mh
- 5 of 15-
To prevent introduction of this erroneous code sequence, the TASKING Assembler V6.0r3 has been
extended with the CHECKBUS18 control which generates a WARNING in the case the described code
sequence appears. When called from within EDE, the Assembler control CHECKBUS18 is
automatically ’activated’.
ADC.11:
Modifications of ADM field while bit ADST = 0
The A/D converter may unintentionally start one auto scan single conversion sequence when the
following sequence of conditions is true:
(1) the A/D converter has finished a fixed channel single conversion of an analog channel n > 0 (i.e.
contents of ADCON.ADCH = n during this conversion)
(2) the A/D converter is idle (i.e. ADBSY = 0)
(3) then the conversion mode in the ADC Mode Selection field ADM is changed to Auto Scan Single
(ADM = 10b) or Continuous (ADM = 11b) mode without setting bit ADST = 1 with the same
instruction
Under these conditions, the A/D converter will unintentionally start one auto scan single conversion
sequence, beginning with channel n-1, down to channel number 0.
In case the channel number ADCH has been changed before or with the same instruction which
selected the auto scan mode, this channel number has no effect on the unintended auto scan
sequence (i.e. it is not used in this auto scan sequence).
Note:
When a conversion is already in progress, and then the configuration in register ADCON is changed,
- the new conversion mode in ADM is evaluated after the current conversion
- the new channel number in ADCH and new status of bit ADST are evaluated after the current
conversion when a conversion in fixed channel conversion mode is in progress, and after the
current conversion sequence (i.e. after conversion of channel 0) when a conversion in an auto scan
mode is in progress.
In this case, it is a specified operational behaviour that channels n-1 .. 0 are converted when ADM is
changed to an auto scan mode while a fixed channel conversion of channel n is in progress (see e.g.
C167 User’s Manual, V2.0, p16-4)
Workaround:
When an auto scan conversion is to be performed, always start the A/D converter with the same
instruction which sets the configuration in register ADCON.
X9:
Read Access to XPERs in Visible Mode
The data of a read access to an XBUS-Peripheral (XRAM, CAN) in Visible Mode is not driven to the
external bus. PORT0 is tristated during such read accesses.
Note that in Visible Mode PORT1 will drive the address for an access to an XBUS-Peripheral, even
when only a multiplexed external bus is enabled.
Semiconductor Group
Errata Sheet, C167CR-4RM, ES-DA, DA, 1.2, Mh
- 6 of 15-
X12:
P0H spikes after XPER write access and external 8-bit Non-multiplexed bus
When an external 8-bit non-multiplexed bus mode is selected and P0H is used for general purpose I/O,
and an internal (byte or word) write access to an XBUS peripheral (e.g. XRAM, CAN, or I²C module) is
performed, and an external bus cycle is directly following the internal XBUS write cycle, then P0H is
actively driven with the write data for approx. 7ns (spikes on P0H).
The spikes also occur if P0H is configured as input. However, read operations from P0H are not
affected and will always return the correct logical state.
The spikes have the following position and shape in a typical application:
spikes occur after the rising edge of CLKOUT which follows the rising edge of ALE for the external bus
cycle
P0H.x = low
--> output low voltage rises to approx. 2.5V, spike width approx. 7ns (@ 0.2 Vcc)
P0H.x = high --> output high voltage drops to approx. 2.0V, spike width approx. 7ns (@ 0.8 Vcc)
Referring to a worst case simulation the maximum width of the spikes may be 15ns with full amplitude
(Vcc/Vss). But this might not be seen on application level.
Note that if any of the other bus modes is selected in addition to the 8-bit non-multiplexed mode, P0H
can not be used for I/O per default.
Workaround:
-
use a different port instead of P0H for I/O when (only) an external 8-bit non-multiplexed bus mode
is selected
or use a different bus type (e.g. 8-bit multiplexed, where P1H may be used for I/O instead of P0H)
or the spikes on P0H may be filtered with an application specific RC element,
or do not perform an external bus access directly after an XBUS write access:
this may be achieved by an instruction sequence which is executed in internal ROM/Flash/OTP,
or internal RAM, or internal XRAM
e.g.
ATOMIC #3
; to prevent PEC transfers which may access external memory
instruction which writes to XBUS peripheral
NOP
NOP
PINS.1:
Output Signal Rise Time
When a non-multiplexed external bus without memory tristate waitstate is used, and an external
access (A) has been performed to a memory location whose address contains many 0s, followed by a
memory access (B) to a location whose address contains many 1s, the following problem may occur:
1. the rising edge of RD# or WR# which is generated at the end of access (A) is relatively slow
2. the rising edge of ALE which is generated at the beginning of access (B) is relatively slow
In an application, we see the following potential problem, in particular at low temperatures:
Timing t28/Address Hold after WR# can no longer be guaranteed, since the address lines (P4, P1)
may already fall below the high reference level before the WR# signal has reached its inactive high
signal level. The high reference level used for timing measurements is 0.2Vcc+0.9 V (= 1.9V @
Vcc=5.0V).
Depending on the characteristic (e.g. input threshold) of an external peripheral (e.g. SRAM), write
cycles may unintentionally overwrite external data.
For RD# cycles, there is no problem since data are always already latched internally by the controller
before the RD# signal is externally driven high.
Workaround:
use a memory tristate waitstate (BUSCONx.MTTCx = 0) for all BUSCON registers which are used for
external write cycles.
Semiconductor Group
Errata Sheet, C167CR-4RM, ES-DA, DA, 1.2, Mh
- 7 of 15-
Note on Interrupt Register behaviour of the CAN module
Due to the internal state machine of the CAN module, a specific delay has to be considered between
resetting INTPND and reading the updated value of INTID. See Application Note AP2924 " Interrupt
Register behaviour of the CAN module in Siemens 16-bit Microcontrollers" on
http://www.siemens.de/semiconductor/products/ics/34/pdf/ap292401.pdf
Deviation from Electrical- and Timing Specification:
The following table lists the deviations of the DC/AC characteristics from the specification in the
C167CR-4RM Data Sheet 7.97 and C167SR/CR-L25M Data Sheet Addendum 1998-03
3UREOHP
3DUDPHWHU
6\PERO
VKRUW QDPH
/LPLW 9DOXHV 8QLW 7HVW
min.
&RQGLWLRQ
max.
DC.IALEH.1
ALE active current
IALEH
1000
instead of
500
-
µA
VOUT = 2.4 V
DC.IRWL.1
RD#/WR# active current
IRWL
-600
instead of
-500
-
µA
VOUT = VOLmax
DC.IP6L.1
Port 6 active current
IP6L
-600
instead of
-500
-
µA
VOUT = VOLmax
DC.IP0L.1
Port 0 configuration current IP0L
-110
instead of
- 100
-
µA
VOUT = VOLmax
Semiconductor Group
Errata Sheet, C167CR-4RM, ES-DA, DA, 1.2, Mh
- 8 of 15-
Problem
Parameter
short name
Symbol
CPU Clock
= 20 MHz
Variable CPU Clock Unit
1/2TCL = 1 to 20 MHz
min.
max.
min.
max.
AC.t5.1
ALE high time
t5
10+ta
instead of
15+ta
-
TCL-15+ta
instead of
TCL-10+ta
-
ns
AC.t12.1
WR#/WRH# low
time
(with RW-delay)
t12
38+tc
instead of
40+tc
-
2TCL-12+tc
instead of
2TCL -10+tc
-
ns
AC.t13.1
WR#/WRH# low
time
(no RW-delay)
t13
63+tc
instead of
65+tc
-
3TCL-12+tc
instead of
3TCL -10+tc
-
ns
AC.t38.1
ALE falling edge to
CS#
t38
-7-ta
instead of
-4-ta
10-ta
-7-ta
instead of
-4-ta
10-ta
ns
AC.t48.1
RDCS#/WRCS# low t48
time
(with RW-delay)
38+tc
instead of
40+tc
-
2TCL-12+tc
instead of
2TCL -10+tc
-
ns
AC.t49.1
RDCS#/WRCS# low t49
time
(without RW-delay)
63+tc
instead of
65+tc
-
3TCL-12+tc
instead of
3TCL -10+tc
-
ns
Notes:
1) Pin READY# has an internal pull-up (all C167xx derivatives). This will be documented in the next
revision of the Data Sheet.
2) Timing t28: Parameter description and test changed from ’Address hold after RD#/WR#’ to ’Address
hold after WR#’. It is guaranteed by design that read data are internally latched by the controller before
the address changes.
3) During reset, the internal pull-ups on P6.[4:0] are active, independent whether the respective pins
are used for CS# function after reset or not.
Semiconductor Group
Errata Sheet, C167CR-4RM, ES-DA, DA, 1.2, Mh
- 9 of 15-
- A/D Converter Characteristics:
ADCC.2.2:
ADC Overload Current
During exceptional conditions in the application system an overload current IOV can occur on the analog
inputs of the A/D converter when VAIN > Vdd or VAIN < Vss. For this case, the following conditions are
specified in the Data Sheet:
IOVmax = | ±5 mA |
The specified total unadjusted error TUEmax = | ±2 LSB | is only guaranteed if overload conditions
occur on maximum 2 not selected analog input pins and the absolute sum of input overload currents
on all analog input pins does not exceed 10 mA. (It is also allowed to distribute the overload to more
than 2 not selected analog input pins).
Due to an internal problem, the specified TUE value is only met for a SRVLWLYH overload current 0 mA ≤
IOV ≤ +5 mA (all currents flowing into the microcontroller are defined as positive and all currents flowing
out of it are defined as negative).
If the exceptional conditions in the application system cause a QHJDWLYH overload current, then the
maximum TUE can be exceeded (depending on value of IOV, RAIN, RAREF and RAGND):
3UREOHP'HVFULSWLRQLQ'HWDLO
2YHUORDG&XUUHQWDWDQDORJ&KDQQHO$1$1DQG$1$1
If a QHJDWLYH overload current IOV occurs on analog input channel ANn (n≠10) then an additional
current IAN (crosstalk current) is caused at the neighbour channels ANn-1 and ANn+1. This
behavior causes an additional unadjusted error AUE to the ADC result.
Relation between IAN and IOV:
IANn+1 = ovf_1 * IOVn
IANn-1 = ovf_1 * IOVn
(n≠10)
(n≠10)
2YHUORDG&XUUHQWDWGLJLWDO&KDQQHO3
A negative overload current IOV at digital Port P7.7 causes an additional current IAN at the analog
input AN0. The relation between both channels is also defined by ovf_1:
IAN0 = ovf_1 * IOVP7.7
2YHUORDG&XUUHQWDWDQDORJ&KDQQHO$1DQG,QIOXHQFHWR9$5()
If an overload current IOV occurs on analog input channel AN10 then an additional current IAREF
(crosstalk current) is caused at pin VAREF.
Depending on RAREF, the internal resistance of the reference voltage, the crosstalk current IAREF at
pin VAREF can cause an additional unadjusted error AUE to all other analog channels.
In case RAREF ≤ 195 Ohm [ RAREF ≤ ((LSB/2) / (IOVmax * ovf_3) ] the maximum possible additional
error to all other channels is smaller than 0.5 LSB with the condition of IOVmax = | ±5 mA| at AN10.
Relation between IAREF and IOV at AN10:
IAREF = ovf_3 * IOV10
(IOV10 : overload current at AN10)
1RWH The influence to the reference voltage VAREF caused by IOV10 (shift of VAREF) is maximum for
VAINn = VAREF and the influence is minimum for VAINn = 0V (n ≠10). The condition RAREF ≤ 195
Ohm and 0.5 LSB is calculated for the worst case at VAINn = VAREF.
Semiconductor Group
Errata Sheet, C167CR-4RM, ES-DA, DA, 1.2, Mh
- 10 of 15-
2YHUORDG&XUUHQWDWDQDORJ&KDQQHO$1DQG,QIOXHQFHWR9$*1'
If an overload current IOV occurs on analog input channel AN10 then an additional current IAGND
(crosstalk current) is caused at pin VAGND.
Depending on RAGND, the resistance between VAGND pin of the microcontroller and analog ground
of the system, the crosstalk current IAGND at pin VAGND can cause an additional unadjusted error
AUE to all other analog channels. In case RAGND ≤ 72 Ohm [ RAGND ≤ ((LSB/2) / (IOVmax * ovf_2) ]
the possible additional error to all other channels is smaller than 0.5 LSB with the condition of
IOvmax = | ±5 mA | at AN10.
Relation between IAGND and IOV10:
IAGND = ovf_2 * IOV10
(IOV10 : overload current at AN10)
1RWH The influence to the reference voltage caused by IOV10 (shifting the potential of VAGND
relative to system ground) is maximum for VAINn = 0V and the influence is minimum for
VAINn = VAREF (n ≠10). The condition RAGND ≤ 72 Ohm and 0.5 LSB is calculated for the
worst case at VAINn = 0V. In standard systems the typical value for
RAGND = 0.1Ohm. In that case the ground shift error is QHJOLJLEOH!
2YHUORDG&XUUHQWDWDQDORJ&KDQQHO$13DQG,QIOXHQFHWR$1
If RAGND ≤ 72 Ohm and RAREF ≤ 195 Ohm then the influence of a negative overload current IOV to
the analog input channel AN11 is:
IAN11 = ovf_1 * IOV10
9DOXHVRIRYIBRYIBDQGRYIB
Parameter
Overload factor_1
Overload factor_2
Overload factor_3
Symbol
ovf_1
ovf_2
ovf_3
Min
- 0.0034
- 0.0068
- 0.0025
Max
0
0
0
(IIHFWVRQWKH&RQYHUVLRQ5HVXOWDQG78(
The effect on the conversion result and the TUE has to be calculated based on the crosstalk
current and the impedance of the analog source RASRC. IANn causes an external voltage U∆n at the
analog channel ANn (same principle for VAREF and VAGND) which is the reason for an additional
unadjusted error $8( of the conversion result. This AUE can increase the specified total
unadjusted error TUE (Specified value: TUE = ± 2 LSB). The voltage U∆n is nearly independent on
the voltage value of the analog source.
U∆n
U∆REF
U∆GND
AUE
TUE
=
=
=
=
=
IANn
* RASRC
IAREF
* RAREF
IAGND * RAGND
U∆n / 1 LSB
(± 2 LSB) ± AUE
[U∆n in mV and LSB in mV]
1RWHA negative overload current IOVn decreases the analog signal voltage VAINn (n≠10) and
causes a negative AUE.
A negative overload current IOV10 (@ AN10) decreases the absolute value of VAREF and
causes a positive AUE.
Semiconductor Group
Errata Sheet, C167CR-4RM, ES-DA, DA, 1.2, Mh
- 11 of 15-
&DOFXODWLRQ([DPSOH
Assumed system values:
= -300 µA
negative overload current at P5.4
IOV4
resistance of the external sensor at P5.5
RASRC = 20 kOhm
1 LSB = 4.9 mV
VAREF = 5 V
RAREF and RAGND has only to be considered for IOV at AN10
Ð GND shift error is negligible
RAGND = 0.1 Ohm
Ð VAREF shift error is negligible
RAREF = 500 Ohm
IAN5
IAN5
IAN5
= ovf_1 * IOV4
= (-3.4*E-3) * (- 300 µA)
= 1.02 µA
U∆5
U∆5
U∆5
= IAN5 * RASRC
= 1.02 µA * 20 kOhm
= 20.4 mV
AUE
AUE
AUE
= U∆n / 1 LSB
= 20.4 mV / 4.9 mV
= 4.2 LSB
5HVXOWThe negative overload current IOV4 of this system example can distort the real result of
AN5 by an additional unadjusted error, -4.2 LSB ≤ AUE ≤ 0 LSB.
The TUE is in the range of -6.2 LSB ≤ TUE ≤ 2 LSB.
Semiconductor Group
Errata Sheet, C167CR-4RM, ES-DA, DA, 1.2, Mh
- 12 of 15-
History List (since device step AB)
Functional Problems
Functional
Problem
Short Description
PWRDN.1
Execution of PWRDN Instruction while pin NMI# = high
CPU.8
Jump instruction in EXTEND sequence
AC
CPU.9
PEC Transfers during instruction execution from Internal RAM
AC
CPU.11
Stack Underflow during Restart of Interrupted Multiply
AC
CPU.16
Data read access with MOVB [Rn], mem instruction to internal ROM
CPU.17
Arithmetic Overflow by DIVLU instruction
BUS.17
Spikes on CS# Lines after access with RDCS# and/or WRCS#
BUS.18
PEC Transfers after JMPR Instruction
RST.1
System Configuration via P0L.0 during Software/Watchdog Timer Reset
AC
RST.3
Bidirectional Hardware Reset
DA
ADC.8
CC31/ADC Interference
AC
ADC.10
Start of Standard Conversion at End of Injected Conversion
AC
ADC.11
Modifications of ADM field while bit ADST = 0
X9
Read Access to XPERs in Visible Mode
X12
P0H spikes after XPER write access and external 8-bit Non-multiplexed bus
PINS.1
OUTPUT Signal Rise Time (DA-step only)
Semiconductor Group
Errata Sheet, C167CR-4RM, ES-DA, DA, 1.2, Mh
Fixed in
step
- 13 of 15-
AC/DC Deviations
AC/DC
Deviation
Short Description
Fixed in
step
DC.VOL.1
Output low voltage (Port0/1/4, ALE, RD#, WR#, ...) test condition 1.6mA (
AC-step only)
DA
DC.IALEH.1
ALE active current 1000µA
DC.IRWL.1 RD#/WR# active current –600µA
DC.IP6L.1
Port 6 active current –600 µA
DC.IP0L.1
Port 0 configuration current –110µA
DC.HYS.1
Input Hysteresis 300mV(restriction not effective in production test)
AC.t5.1
ALE high time TCL-15ns
AC.t12.1
WR#/WRH# low time (with RW-delay) 2TCL-12ns
AC.t13.1
WR#/WRH# low time (no RW-delay) 3TCL-12ns
AC.t38.1
ALE falling edge to CS# -7ns
AC.t48.1
RDCS#/WRCS# low time (with RW-delay) 2TCL-12ns
AC.t49.1
RDCS#/WRCS# low time (no RW-delay) 3TCL-12ns
ADCC.2.2
ADC Overload Current
1)
Note 1): problem not included in AC-step
Semiconductor Group
Errata Sheet, C167CR-4RM, ES-DA, DA, 1.2, Mh
- 14 of 15-
In addition to the description in the C167 Derivatives User’s Manual V2.0, the following feature
enhancements have been implemented in the C167CR-4RM:
Incremental position sensor interface
For each of the GPT1 timers T2, T3, T4 of the GPT1 unit, an additional operating mode has been
implemented which allows to interface to incremental position sensors (A, B, Top0). This mode is
selected for a timer Tx via TxM = 110b in register TxCON, x = (2, 3, 4). Optionally, the contents of T5
may be captured into register CAPREL upon an event on T3. This feature is selected via bit CT3 = 1 in
register T5CON.10
Compatibility with previous versions:
In previous versions (e.g. C167CR-LM BA-step), both of the settings (TxM = 110b, T5CON.10 = 1)
were reserved and should not be used. Therefore, systems designed for previous versions will also
work without problems with the C167CR-4RM.
Oscillator Watchdog
The C167CR-4RM provides an Oscillator Watchdog (OWD) which monitors the clock at XTAL1 in
direct drive mode. In case of clock failure, the PLL Unlock/OWD Interrupt Request Flag (XP3IR) is set
and the internal CPU clock is supplied with the PLL basic frequency. This feature can be disabled by a
low level on pin Vpp/OWE. See also C167CR-4RM Data Sheet 7.97.
Bidirectional Reset
The C167CR-4RM allows to indicate an internal watchdog timer or software reset on the RSTIN# pin
which will be driven low for the duration of the internal reset sequence. This option is selectable by
software via bit BDRSTEN/SYSCON.3. After reset, the bidirectional reset option is disabled
(BDRSTEN/SYSCON.3 = 0). See also C167CR-4RM Data Sheet 7.97. Beginning with the AC-step of
the C167CR-4RM, RSTIN# will also be driven low for the duration of the internal reset sequence when
this reset was initiated by an external HW reset signal on pin RSTIN#.
Please note also the following functional difference to the C167CR-LM BA-step:
XBUS Peripheral Enable Bit XPEN/SYSCON.2
In the C167CR-4RM, bit SYSCON.2 is a general XBUS Peripheral Enable bit, i.e. it controls both the
XRAM and the CAN module.
Compatibility with previous versions:
When bit SYSCON.2 = 0 (default after reset) in the C167CR-4RM, and an access to an address in the
range EF00h ... EFFFh is made, either an external bus access is performed (if an external bus is
enabled), or the Illegal Bus Trap is entered. In previous versions (e.g. C167CR-LM BA-step), the CAN
module was accessed in this case.
Systems where bit SYSCON.2 was set to ’1’ before an access to the CAN module in the address range
EF00h ... EFFFh was made will also work without problems with the C167CR-4RM.
Application Support Group, Munich
Semiconductor Group
Errata Sheet, C167CR-4RM, ES-DA, DA, 1.2, Mh
- 15 of 15-