コミット: abb5d611bd9536cb076b1904b9f947c954725189 - jmr6805break (git)

およそ20年前に、68HC05 の開発の練習に書いた車のブレイクライト・方向指示器コントローラです。

コミットメタ情報

• リビジョン: abb5d611bd9536cb076b1904b9f947c954725189 (tree)
• 日時: 2013-07-08 21:19:51
• 作者: Joel Matthew Rees <reiisi@user...>
• コミッター: Joel Matthew Rees

ログメッセージ

Improvements in the debounce and operational analysis.
(The previous commit added break light and related functionality.)

変更サマリ

• delete: TURNSIG.AS1 => TURNSIGQ.AS2

差分

--- a/TURNSIG.AS1
+++ b/TURNSIGQ.AS2
@@ -71,16 +71,14 @@
 * feedback and dash panel indicators should be sufficient user feedback.
 
 *       The turn signal return mechanism is a pair of reed relays
-* mounted inside the bottom of the steering column cowling about ten
-* degrees apart, with three magnetic actuators mounted to the steering
-* shaft at ten o'clock, two o'clock, and six o'clock.  (More magnets and
-* other patterns might also be workable, but they should be tested in
-* real steering assembly before the design is locked in.  The shape and
-* position of the magnets and reeds should be such that the reeds close
-* only once as a magnet passes, and such that the reeds have about ten
-* degrees of overlapping actuation. The overlap in actuation helps
-* determine the direction the steering wheel is rotating.  Minor changes
-* to the design should allow the use of optical sensors, but I have the
+* mounted inside the bottom of the steering column housing about 11.25
+* degrees apart, with three to eight magnetic actuators distributed
+* around the steering shaft.  The shape, size, and position of the
+* magnets and reeds should be such that the reeds close only once as a
+* magnet passes, and such that the reeds have about 22.5 degrees of
+* overlapping actuation. The overlap in actuation is used to determine
+* the direction the steering wheel is rotating.  Minor changes to the
+* design should allow the use of optical sensors, but I have the
 * impression that magnetic devices are more immune to extremes of
 * climate.
 
@@ -173,35 +171,60 @@
 * not naturally biased, so the the emergency flasher seemed a more
 * natural match for IRQ.
 
-*       The right turn signal return switch is positioned about five
-* degrees right of bottom of the steering column housing, and the left
-* return is positioned the same distance to the left.  In this way, when
-* the steering wheel rotates, the return switch for the direction of
-* rotation actuates first, then the opposite direction switch closes,
+*       The following crude diagram illustrates the actuation in a left
+* rotation (restoring from a right turn or turning left).  The carots
+* indicate the position of the reed switches, the asterisks indicate the
+* actuation field of a magnet.
+
+*    Actuation Field ********
+*         Left Return Switch ^  ^ Right Return Switch
+*                      ********
+*                            ^  ^
+*                        ********
+*                            ^  ^
+*                          ********
+*                            ^  ^
+*                            ********
+*                            ^  ^
+*                              ********
+*                            ^  ^
+*                                ********
+*                            ^  ^
+
+*       Referring to the above diagram, the right turn signal return
+* switch is positioned about 5 5/8 degrees right of bottom center below
+* the steering column and fixed to the steering column housing, and the
+* left return is positioned the same distance to the left.  In this way,
+* when the steering wheel rotates, the return switch for the direction
+* of rotation actuates first, then the opposite direction switch closes,
 * then the same direction switch opens, and then the opposite direction
 * switch opens.  Thus, when restoring from a turn, the return switch for
-* the appropriate direction closes after the other switch.  In other
-* words, if you turn right, the steering wheel will be rotate to the
-* left when you return to straight, and the switches actuate from left
-* to right during the restore.
-
-*       When both turn signals are off, both turn signal ports are set
-* as inputs.  When a turn signal switch input senses a closure, that
-* port is made an output and driven high.  When the cancel input senses
-* a closure, the turn signal ports are restored to both input.
-
-*       If the cancel input remains high, a manual cancel is decoded.
-* If not, the other direction input is made output and driven high.  If
-* the cancel input returns high now, the other direction switch closed
-* first, and a restoring rotation is decoded.  If not, same direction
-* rotation is decoded; then the the opposite direction port is returned
-* to input, and the original direction port is again made output and
-* driven high to wait again.
-
-*       In the meantime, if the opposite direction turn signal input
-* senses a closure, then it is assumed that the operator changed his
-* mind without hitting the manual cancel, and the roles of the turn
-* signal I/O bits are reversed.
+* the direction of the turn closes after the other switch.  In other
+* words, if you turn right, the steering wheel will rotate to the left
+* when you return to straight, and the switches actuate from left to
+* right during the restore.  (I say this so many times because it takes
+* repeating to convince myself of it.)
+
+*       When both turn signals are off, both turn signal switch ports
+* are set as inputs.  When a turn signal switch input senses a closure,
+* its port is made an output and driven high.  Now, in addition to
+* manual cancel, the cancel port can sense steering wheel rotation.  In
+* same direction rotation, there is a lead period when the cancel input
+* does not remain high if the turn signal input port states are
+* inverted. Debounce suppresses cancel under these conditions.
+  
+*       Since the opposite direction turn signal switch port remains an
+* input, it can still be sensed.  The operator can change turn signals
+* directly, without hitting the cancel button.
+
+*       The separation of the switches and the actuation arc length I
+* specify are based on my assumption that the steering wheel will rotate
+* at a maximum speed of three revolutions per second, and that the four
+* to six mS required debounce period requires a minimum 10 milli-second
+* separate actuation period.  The actual result was 10.9 degrees
+* separate actuation (21.8 total), which I arbitrarily rounded up to a
+* power of two fraction of the circle: 1/16 circle total actuation and
+* 1/32 circle separate actuation.
 
 *       A peculiar characteristic of this design is the apparent lack of
 * use for the RESET input.  Since power-on reset is built in to the
@@ -243,6 +266,7 @@ RGT_BT_ equ B0_
 
 * logical input definitions for debounce and states
 * bits are remapped for debounce and queue speed
+* six bits of physical input, high two bits used for cancel logic
 F_LFT   equ LFT_BT
 F_LFT_  equ LFT_BT_
 F_RGT   equ RGT_BT
@@ -255,6 +279,8 @@ F_EMG   equ {DRVDRV < 2}        ; fold onto output bit
 F_EMG_  equ {DRVDRV_ + 2}
 F_CAN   equ {CAN_BT < 2}
 F_CAN_  equ {CAN_BT_ + 2}
+F_OPC   equ $40                 ; opposite direction cancel record
+F_MNC   equ $80                 ; manual cancel record
 
 *       Event queue constants
 * We only have 6 bit events and only record 4 in the queue.
@@ -290,6 +316,8 @@ STATES  rmb 1   ; current controller states
 ENG_TIM rmb 2   ; engine time out timer
 FLASH   rmb 1   ; flash timer
 FLTIME  rmb 1   ; flash flash time to load
+CANCEL  rmb 1   ; debounce for cancel button
+RETURN  rmb 1   ; debounce for return logic
 
 
 * ROM definitions follow:
@@ -297,6 +325,7 @@ FLTIME  rmb 1   ; flash flash time to load
 
         SUBHEADER 'Timer Service Code -- debounce'
         PAGE
+
 *       The timer service routine has five functions: debouncing the
 * inputs, directly setting the brake lights, loading the flasher event
 * queue, and maintaining two software clocks. Interrupts occur on timer
@@ -308,8 +337,15 @@ FLTIME  rmb 1   ; flash flash time to load
 * the debounced input in TOGGLE, and leaves the current debounced input
 * in ISTATE.  Since debouncing occurs on three identical states, the
 * debounce period is between two and three timer overflows -- 4 to 6 mS
-* with a 1 MHz crystal, 2 to 3 mS with a 2 MHz crystal.  Debouncing
-* ignores the return switch logic and related activities.
+* with a 1 MHz crystal, 2 to 3 mS with a 2 MHz crystal.
+
+*       This debounce technique works because we don't care what is
+* happening to the inputs between timer overflows, and because there is
+* no mechanism for noise coincident and rythmic with timer overflow.
+
+*       The cancel TOGGLE bit is suppresed on same-direction rotation.
+* Trailing edge TOGGLEs are filtered out for all but brake and engine-on.
+* Cancel direction and manual cancel TOGGLEs are always filtered out.
 
 $CYCLE_ADDER_ON
 TIMSRV  bset TOFR_,TSCR
@@ -317,19 +353,45 @@ TIMSRV  bset TOFR_,TSCR
         sta DEBO+2
         lda DEBO
         sta DEBO+1
+
 * read the inputs, converting physical to logical
-        lda PORTB       ; PB2-7 always read 0
-        sta DEBO
         lda PORTA
         and #{BRK_SW | ENG_SW | CAN_BT }        ; clears F_EMG
-        bih TIM0MG      ; invert EMG and fold it in
+        bih TMMEMG      ; invert EMG and fold it in
         ora #F_EMG
-TIM0MG  lsra    ; remap
+TMMEMG  lsra    ; remap
         lsra
+        sta DEBO
+        lda PORTB       ; PB2-7 always read 0
         ora DEBO
-        sta DEBO        ; current state
+        sta DEBO
+
+* check for turn signal return and cancel
+        bit #F_CAN      ; nothing at all without cancel
+        beq TMSTGL
+        lda DDRB        ; meaningless when no turn signal
+        beq TMSTGL
+$CYCLE_ADDER_OFF
+$CYCLE_ADDER_ON
+        tax
+        eor #{LFT_BT | RGT_BT}  ; strobe other side
+        sta PORTB       ; set output state first
+        sta DDRB        ; handy output state, huh?
+        brset F_CAN_,PORTA,TM0000       ; F_CAN into carry
+TM0000  rora
+        clr DDRB        ; no strobe: manual cancel?
+        brset F_CAN_,PORTA,TM0001       ; F_CAN into carry again
+TM0001  rora
+        and #{F_OPC | F_MNC}
+        ora DEBO
+        sta DEBO
+        stx PORTB       ; restore turn signal inputs
+        stx DDRB
+$CYCLE_ADDER_OFF
+$CYCLE_ADDER_ON
+
 * debounce inputs, set flags in TOGGLE
-        lda DEBO
+TMSTGL  lda DEBO
         eor DEBO+1
         sta TOGGLE      ; re-use TOGGLE, TOGGLE not valid
         lda DEBO+1
@@ -342,7 +404,7 @@ TIM0MG  lsra    ; remap
         txa
         and ISTATE      ; look only at bits that are stable
         eor STABLE      ; flag bits that toggled
-        sta TOGGLE      ; TOGGLE now valid
+        sta TOGGLE      ; TOGGLE valid, but not filtered
         coma            ; mask of non-toggles
         and ISTATE      ; old state of non-toggles
         sta ISTATE      ; (toggled bits cleared)
@@ -351,8 +413,21 @@ TIM0MG  lsra    ; remap
         ora ISTATE
         sta ISTATE      ; ISTATE now has current debounced inputs
 
+* filter TOGGLE F_CAN if same direction
+        bit #{F_OPC | F_MNC}    ; opposite direction or manual?
+        bne TMFFAL
+        bclr F_CAN_,TOGGLE      ; same direction
+
+* filter falling edge of F_LFT, F_RGT, F_EMG, F_CAN in TOGGLE
+* assumes ISTATE still in A
+TMFFAL  ora #{F_ENG | F_BRK}    ; not filtered
+        and TOGGLE      ; filter any other fall(en) edge
+        and #{$FF ^ F_OPC ^ F_MNC}      ; these should never TOGGLE
+        sta TOGGLE
+
         SUBHEADER 'Timer Service Code -- brake lights'
         PAGE
+
 *       Brake lights are the highest priority function, depending only
 * on the ISTATE F_BRK.  The STATES F_BRK is returned as a flag to other
 * routines.
@@ -370,6 +445,7 @@ TMBRKDUN
 
         SUBHEADER 'Timer Service Code -- fill event queue'
         PAGE
+
 *       General debounce complete; now put signal events in queue:
 * Brake and engine-on are handled in TIMSRV, as synchronous events, and
 * so are not queued.  Turn signal and emergency inputs are positive edge
@@ -381,7 +457,6 @@ TMBRKDUN
 
         lda TOGGLE      ; F_BRK & F_ENG are not queued!
         and #{$FF ^ F_BRK ^ F_ENG}      ; KICS assembler has no bit not?
-        and ISTATE      ; queue only TOGGLE on
         sta TOGGLQ
         beq TMCLOCK     ; save time
         ldx QIN
@@ -504,7 +579,7 @@ RSTSRV  lda #{F_LDRV | F_RDRV | R_LDRV | R_RDRV}
 
 * Falls through to MAINLOOP
 
-        SUBHEADER 'STOP and dequeue'
+        SUBHEADER 'Dequeue Events and Modify STATES, also STOP'
         PAGE
 
 MAINLOOP
@@ -561,20 +636,22 @@ MNQNWRP stx QOUT
 
 MNQCAN  cmp #F_CAN
         bne MNQLFT
-        lda DDRB
-        eor #3
-
-        clr DDRB
-        bsr CANDEBO
-        cmp #6
-        blo MNQCAND
-        bset F_CAN_,STATES
-        bra MNDOSTATES
-MNQCAND lda STATES
+        ldx #7
+
+        lda STATES
         and #{F_LFT | F_RGT}
-        eor #{F_LFT | F_RGT}    ; other port
-        sta DDRB
-        bsr CANDEBO
+        beq MNDOSTATES
+        eor #{F_LFT | F_RGT}
+        sta TMPTURN
+        sta PORTB
+MNQCANL sta DDRB
+        brclr CAN_BT_,PORTA,MNQCANM     ; opposite direction?
+        incx
+MNQCANM clr DDRB        ; all outputs pulled down
+        brclr CAN_BT_,PORTA,MDQCANT     ; manual cancel?
+        incx
+        dec something
+        bne MNQCANL
 
 
 MNQLFT  cmp #F_LFT
@@ -591,21 +668,6 @@ MNQRGT  cmp #F_RGT
 
 MNQDEF  jmp SWISRV      ; nothing else should be queued
 
-* subroutine to quickly "debounce" the cancel input
-* reads CAN_BT eight times, summing the bits into A
-* trashes X
-$CYCLE_ADDER_ON
-CANDEBO ldx #8
-        clra
-        brclr CAN_BT_,PORTA,CANDEB1     ; get CAN_BT into carry
-CANDEB1 adc #0
-        decx
-        bne CANDEBO
-$CYCLE_ADDER_OFF
-* about 1/4 mS at 1 MHz, sufficient to believe the input bit
-        rts
-
-
 * then make the output mirror the state