Can't reliably transfer over serial connection between two PICs

[CuriousOne]

Hello.
Having a strange issue.

PIC18F45K80 and PIC16F1936.

they are connected to each other via 22K resistor. PORTC.3 on both.

Trying to run the simple code:

TX: (1936)



barka:
serout2 portc.3,84,[345]
pause 100
goto barka


RX (45K80)



taka:
serin2 portc.3,84,[x]
if x=345 then stop
lcdout $fe, $c0, "X=", dec x, " "
goto taka


It does not work, it never receives "345". received value is always random, but quite often it is 89.
Variable types are proper, port configs also.
I tried to change direction and send from 18F to 16F - same issue.
Checked signals with scope - no issues.
Tried to various resistors - 1K, 10K, 22K - no difference.

Yes I know that hardware port should be used, but PCB is already made.
Distance from one chip to another is about 10 centimeters and they're both on the same PCB and same power supply. I've checked signals with scope and there are no issues, like low levels or noise.

I tried to make code more complicated, like sending 8 characters + control character (enter) -this improves it a bit, might say that at least one try from 10 delivers proper result. Sometimes, it starts to work fine, but then again messes up. Also tried reducing port speed to 1200bps - no change.


Any ideas?

[Ioannis]

1. How can you be sure that the receiving PIC will start "listening" just before the sending PIC? Better use a start character on the sender and use this character as a WAIT one on the receiving PIC to synchronize both.

2. How is variable x defined? BYTE or WORD? 345 is more than a BYTE can store. 345 in Binary is 101011001. So if x is byte, it will randomly store 01011001 which happens to be, guess what?
Exactly 89...!!!

Ioannis

[CuriousOne]

Yes, X is word type, and it sometimes get values over 8 bits, but never 345.
I tried sending lower numbers, like 4 or 16 or whatever - they also got distorted.

I also tried like this:

arraywrite textline, ["ABCD1234"]serout2 portc.3,84,[STR textline\8,13]

Sometimes whole string is delivered fine, sometimes it is all garbled.

As I understand, you suggest to use WAIT/WAITSTR statements?

Meanwhile, by help of AI, this crude bit-banging "protocol" was developed. By hand tuning PAUSEUS values, it works, but I'd prefer "factory" solution if possible.

Sender:


Main:
FOR TxValue = 0 TO 254
GOSUB SendNumber
PAUSE 300 ' Wait before sending next number
NEXT
GOTO Main


' --- Subroutine: SendNumber ---
SendNumber:
' Start bit (LOW pulse)
LOW DataPin
PAUSEUS 500
HIGH DataPin
PAUSEUS 500


' Send 4 bits (LSB first)
FOR BitMask = 0 TO 7
IF (TxValue >> BitMask) & 1 THEN
HIGH DataPin
ELSE
LOW DataPin
ENDIF
PAUSEUS 500
HIGH DataPin ' Return line HIGH between bits
PAUSEUS 500
NEXT
RETURN


Receiver



Mainx:
' --- Wait for start bit: must see line LOW ---
WaitForStart:
IF DataPin = 1 THEN WaitForStart ' Keep waiting while idle is HIGH
PAUSEUS 490 ' Wait a bit into LOW
IF DataPin = 0 THEN ' Still LOW? good start
GOSUB ReadBits
ENDIF
GOTO Mainx


' --- Subroutine: ReadBits ---
ReadBits:
' Wait until line goes back HIGH before first bit
DO
LOOP UNTIL DataPin = 1
PAUSEUS 1000 ' Move to middle of first bit period


RxValue = 0
FOR BitMask = 0 TO 6
IF DataPin = 1 THEN
RxValue = RxValue | (1 << BitMask)
ENDIF
PAUSEUS 1020 ' Move to middle of next bit slot
NEXT


' --- Ensure line is HIGH again before next frame ---
DO
LOOP UNTIL DataPin = 1


' --- Display received value on LCD ---
LCDOUT $FE,1 ' Clear display
LCDOUT "Received: ", DEC RxValue


RETURN

[CuriousOne]

I just tried this code, and it does not work:


TX:



Main:
FOR TxValue = 0 TO 255

SEROUT2 DataPin, 84, [100,5,200, TxValue]
PAUSE 500
NEXT
GOTO Main


RX:



LCDOUT $FE,1
LCDOUT "Waiting data..."






RxValue VAR BYTE


Mainx:


SERIN2 RxPin, 84, [WAIT (100,5,200), RxValue]


LCDOUT $FE, $C0, "Rx=", dec RXvalue, " "


GOTO Mainx

[CuriousOne]

if I reduce control values to 1 or two, then it works, but totally wrong data is being decoded.

[Ioannis]

How about this code:



x vart word
x=345
barka:
serout2 portc.3,84,["s",x]
pause 100
goto barka


RX (45K80)



x var word
taka:
serin2 portc.3,84,[wait("s"), x]
if x=345 then lcdout $fe, $c0, "X=", dec x, " "
goto taka


Ioannis

[Jerson]

Why 22k? It seems pretty high to have. Your post suggests you've tried a lower value like 1K; if both PICs are running at the same voltage levels, then altogether do away with the resistor. Due to the resistor, it is possible the logic low level is not being met. If you fix this and still get the problem, then continue below.

I suspect the problem could be the software serial not getting the start bit reliably.

What you can do - since you use the 1936 is to configure the receive pin as interrupt on change pin; I believe all pins have IOC ability on this chip.

When you get the IOC interrupt, invoke the serin2 command in the ISR

I cannot advise with code as I am not up-to-date with PicBasic since some time now.

[CuriousOne]

Both 1K and 22K are actually suggested in PBP manual.
So I tried both, I tried 100 ohm resistor, I tried direct connection. No practical difference - actual "junk" bytes are getting different, but proper one still not delivered. The above slow code, suggested by AI, works with either 100 ohm, 1K or 22K.

Currently I'm sending data from 1936 to K80 because K80 has LCD screen so I can use it for debugging. In the real life, K80 will be sending data to 1936.

As a side "improvement", I can use additional data line from K80 to 1836, so SHIFTOUT/SHIFTIN should work then. I guess it will be more reliable?

[Ioannis]

The 22K is used ONLY when you are driving the INPUT of the PIC from a real RS232 line, which does -12 to +12 Volts on the pulses!

No need to have it on TTL connection as in your case. Remove it.

The 1K is used ONLY when you drive from the OUTPUT of the PIC to the RS232 line, to lower the current that PIC is forced to supply.

Again no need to have it on TTL connection as in your case. Remove it.

Have you tested the code on #6 with no resistors?

Ioannis

[richard]

Have you tested the code on #6 with no resistors?

works ok in simulator, apart from initial sync issue due to the comms line not beginning at a proper tty level before reception enabled

and then again i used complete programs not snippets full of guess work


@Ioannis you missed the spam in faq section

[CuriousOne]

I haven't tried that code yet, will try it today later.

Regarding "other" code, there is none, except config and variable assignments.

Here's complete code for 16F1936 before I launch that part of code:



;----[16F1937 Hardware Configuration]-------------------------------------------
#IF __PROCESSOR__ = "16F1936"
#DEFINE MCU_FOUND 1
#CONFIG
cfg1 = _FOSC_INTOSC ; INTOSC oscillator: I/O function on CLKIN pin
cfg1&= _WDTE_OFF ; WDT disabled
cfg1&= _PWRTE_OFF ; PWRT disabled
cfg1&= _MCLRE_OFF ; MCLR/VPP pin function is digital input
cfg1&= _CP_ON ; Program memory code protection is enabled
cfg1&= _CPD_OFF ; Data memory code protection is disabled
cfg1&= _BOREN_OFF ; Brown-out Reset disabled
cfg1&= _CLKOUTEN_OFF ; CLKOUT function is disabled. I/O or oscillator function on the CLKOUT pin
cfg1&= _IESO_ON ; Internal/External Switchover mode is enabled
cfg1&= _FCMEN_ON ; Fail-Safe Clock Monitor is enabled
__CONFIG _CONFIG1, cfg1


cfg2 = _WRT_OFF ; Write protection off
cfg2&= _VCAPEN_OFF ; All VCAP pin functionality is disabled
cfg2&= _PLLEN_OFF ; 4x PLL disabled
cfg2&= _STVREN_ON ; Stack Overflow or Underflow will cause a Reset
cfg2&= _BORV_19 ; Brown-out Reset Voltage (Vbor), low trip point selected.
cfg2&= _LVP_OFF ; High-voltage on MCLR/VPP must be used for programming
__CONFIG _CONFIG2, cfg2


#ENDCONFIG


#ENDIF


;----[Verify Configs have been specified for Selected Processor]----------------
; Note: Only include this routine once, after all #CONFIG blocks
#IFNDEF MCU_FOUND
#ERROR "No CONFIGs found for [" + __PROCESSOR__ +"]"
#ENDIF


'nixie led clock electronics direct drive





ADCON1=%11110011 'enable adc internal reference and justify
FVRCON=%11001111 'set internal reference to 4.096V
OSCCON=%11110000 'SET INTOSC TO 8MHZ
'ANSELE=%00000000 'enable adc input porte.2
ANSELA=%00000000 'disable ADC on A
ANSELB=%00000000 'disable ADC on B
'ANSELD=%00000000 'disable ADC on D
TRISC=%00000000 'set PORTC as output all
'TRISD=%00000000 'set PORTD as output all
TRISB=%00000000 'set PORTB as output all
TRISA=%00000000 'set PORTA 2 as input, others as output
TRISE=%00000000 'set PORTE as output all
WPUB=%00000000 'DISABLE B PULL UPS
OPTION_REG=%100000000
LCDCON=%00000000 'disable LCD controller pins
LCDSE0=%00000000
LCDSE1=%00000000
'LCDSE2=%00000000


DEFINE OSC 32 'set oscillator speed




;____[ For 12F/16F only - Interrupt Context save locations]_________________
wsave var byte $20 SYSTEM ' location for W if in bank0
'wsave var byte $70 SYSTEM ' Alternate save location for W
' if using $70, comment out wsave1-3
' --- IF any of these next three lines cause an error ?? -------------------
' Comment them out to fix the problem ----
' -- The chip being used determines which variables are needed -------------
wsave1 VAR BYTE $A0 SYSTEM ' location for W if in bank1
wsave2 VAR BYTE $120 SYSTEM ' location for W if in bank2
wsave3 VAR BYTE $1A0 SYSTEM ' location for W if in bank3
'---DO NOT change these-----------------------------------------------------
ssave VAR BYTE BANK0 SYSTEM ' location for STATUS register
psave VAR BYTE BANK0 SYSTEM ' location for PCLATH register


;----[ MIBAM Setup ]--------------------------------------------------------
BAM_COUNT CON 13 ; How many BAM Pins are used?
INCLUDE "MIBAM.pbp" ; Mirror Image BAM module




dta var porta.6 'LED PINS
clk var porta.7
FET var portc.2 'current control fet pin


VALUE VAR BYTE 'DIGIT DECODER VALUE VARIABLE
LOLD VAR BYTE 'OLD LEFT VALUE FOR DECODER
ROLD VAR BYTE 'OLD RIGHT VALUE FOR DECODER
'LED COLOR VALUES 'LEFT AND RIGHT
LR VAR BYTE
LG VAR BYTE
LB VAR BYTE
LX VAR BYTE 'LEFT BRIGHTNESS
PACKED VAR BYTE 'LED PACKER
'
RR VAR BYTE
RG VAR BYTE
RB VAR BYTE
RX VAR BYTE 'RIGHT BRIGHTNESS
new var byte 'current display value
old var byte 'old display value
maxbr var byte 'brightness holder
CURBR VAR BYTE 'CURRENT RELATIVE BRIGHTNESS HOLDER
ANIT var byte 'animation type digits
DOTA var byte 'dot animation type
x var word ' main small counter
y var word 'additional var
BR1 var byte 'brightness comparator 1
BR2 var byte '2
DCMODE var byte '6 or 10 digit mode selector






tmp var byte 'temporal var for counting
TEXTLINE VAR BYTE[8] 'MAIN ARRAY FOR TEXT STORAGE


'DIGIT PINS
D0 VAR BYTE
D1 VAR BYTE
D2 VAR BYTE
D3 VAR BYTE
D4 VAR BYTE
D5 VAR BYTE
D6 VAR BYTE
D7 VAR BYTE
D8 VAR BYTE
D9 VAR BYTE
DF VAR BYTE 'TEMPERATURE SIGN
DT1 VAR BYTE 'DOTS
DT2 VAR BYTE


eval var byte 'eeprom reader
ELOC VAR BYTE 'EEPROM LOCATION


ASM
BAM_LIST macro ; Define PIN's to use for BAM


BAM_PIN (PORTB,3, D0)
BAM_PIN (PORTA,0, D1)
BAM_PIN (PORTB,4, D2)
BAM_PIN (PORTC,4, D3)
BAM_PIN (PORTB,5, D4)
BAM_PIN (PORTC,7, D5)
BAM_PIN (PORTB,2, D6)
BAM_PIN (PORTC,5, D7)
BAM_PIN (PORTB,1, D8)
BAM_PIN (PORTB,0, D9)
BAM_PIN (PORTC,6, DF)
BAM_PIN (PORTA,3, DT1)
BAM_PIN (PORTA,2, DT2)

endm
BAM_INIT BAM_LIST ; Initialize the Pins
ENDASM

[richard]

And there is the answer, software timed serial reception and transmission will be mangled by the mibam interrupt
Proof yet again of the total worthlessness of snippets

[CuriousOne]

Don't think so.

Generated serial data is perfect (checked with scope), and my scope, in logic analyzer mode, easily and properly decodes data sent by 1936. Also, I've generated couple of PAUSE and PAUSEUS statements and they all seem to be working fine in terms of timing.

[Ioannis]

I am pretty sure that Richard is absolutely correct about MiBAM interrupts. Since they run all the time, surely will mess with other tasks. Maybe your scope did not trigger on the proper time.

@Richard. Did not quite follow thre FAQ thing...

Ioannis

[richard]

In FAQ the latest post "how to use code tags" is spam, cannot link it using phone as browser

[CuriousOne]

So since I can't get rid of MIBAM (it is essential) this means that I have to use that software protocol provided by AI?

[tumbleweed]

In general, ANY bit-banged protocol that's generated in software, whether it's SHIFTIN/SHIFTOUT/SERIN/SEROUT is going to be unreliable when used with interrupts running. That's one reason hardware peripherals exist.

A software-based receiver is notoriously unreliable even without interrupts since there's no buffering at all.

You can get simple examples and tests to work, but they'll be extremely fragile.

[CuriousOne]

Well, my use case is quite different.
The receiver (1936) which runs MIBAM, does nothing besides BAMing and checking serial input, if anything is delivered.
"Anything" is delivered in form of single byte, once per second, minute or hour.

I'll try to verify the code without MIBAM and see whenever it causes the issues.

And if I switch to hardware serial input, will it still work with MIBAM?

[richard]

Well, my use case is quite different.


Well, my use case is ̶q̶u̶i̶t̶e̶ ̶d̶i̶f̶f̶e̶r̶e̶n̶t̶.̶ exactly that. ie. bit-banged protocol that has its timing generated in software and is going to be unreliable when used with interrupts running

[Ioannis]

And if I switch to hardware serial input, will it still work with MIBAM?

Yes it will work because it has a two byte buffer and does not rely on software to catch that incoming byte.

@ Richard: I got it. Hope it is OK now. The spammer hit the road. I don't know why this was missed from all of thw mods...

Ioannis

[CuriousOne]

So I have 6 slave 1936 and 1 master K80.
Now they are all connected via independent pins to K80, where serial transfer goes now
And there is also one common pin, which connects them altogether (not used right now)
So my idea is, to use that common pin as serial data pin, and use these independent pins as enable input for that particular MCU ?

[Ioannis]

Or use the open output feature and tie all together on one pin.

Ioannis

[CuriousOne]

Sorry I had a little time to check all suggestions, but if there's a timing issue, I have series of 16 APA102C leds connected to the same 1936, they are driven via SHIFTOUT and everything works properly. So if MIBAM is messing with SERIN2, it should mess with SHIFTOUT as well?

[amgen]

just curious.... CuriousOne.......
how are you looping program ???
For timing critical and ser-in's, interrupts are best for not missing incoming stuff.
So how did you set up your receiver routines?

[richard]

I have series of 16 APA102C leds connected to the same 1936is that the same 1936 that

The receiver (1936) which runs MIBAM, does nothing besides BAMing and checking serial input, if anything is delivered.
"Anything" is delivered in form of single byte, once per second, minute or hour.


So if MIBAM is messing with SERIN2, it should mess with SHIFTOUT as well?

sort of, the timing will be screwed up but since the pic is master and clocking is synchronous it won't matter the apa102c's are slaved to the masters clock.
pbp has shiftin and shiftout commands but they both act as master, there is no shiftin slave mode command.
any bit-banged reception code you make would need to honour the synchronous nature of the protocol i.e. interrupts will be a problem

there is a reason hardware peripherals exist.

[CuriousOne]

Yes, it is same 1936 which is running MIBAM.
So to check theory about whenever MIBAM causes issues to serial timings, I've connected 16 led strip (longest I had at moment) with APA102Cs to it, and mibam code was running in background and I was shifting out data without any issues.

[richard]

So to check theory about whenever MIBAM causes issues to serial timings, I've connected 16 led strip (longest I had at moment) with APA102Cs to it, and mibam code was running in background and I was shifting out data without any issues.


its not a valid test, it means nothing. shiftout clock timing for a slave is synchronous , serin serout are asynchronous
pbp has no synchronous reception commands as a slave device

[tumbleweed]

"I didn't see the problem, therefore it doesn't exist "

Famous last words. Just ask Murphy.

[CuriousOne]

I did not went that far so far :)
Will do more tests this weekend.

[CuriousOne]

Did a detailed analysis of how MIBAM affects pause and general code execution at PIC16F1936 running at 32mhz and with 13 BAM outputs.

This is the simple code, and output is monitored with scope.



;----[16F1937 Hardware Configuration]-------------------------------------------
#IF __PROCESSOR__ = "16F1936"
#DEFINE MCU_FOUND 1
#CONFIG
cfg1 = _FOSC_INTOSC ; INTOSC oscillator: I/O function on CLKIN pin
cfg1&= _WDTE_OFF ; WDT disabled
cfg1&= _PWRTE_OFF ; PWRT disabled
cfg1&= _MCLRE_OFF ; MCLR/VPP pin function is digital input
cfg1&= _CP_ON ; Program memory code protection is enabled
cfg1&= _CPD_OFF ; Data memory code protection is disabled
cfg1&= _BOREN_OFF ; Brown-out Reset disabled
cfg1&= _CLKOUTEN_OFF ; CLKOUT function is disabled. I/O or oscillator function on the CLKOUT pin
cfg1&= _IESO_ON ; Internal/External Switchover mode is enabled
cfg1&= _FCMEN_ON ; Fail-Safe Clock Monitor is enabled
__CONFIG _CONFIG1, cfg1


cfg2 = _WRT_OFF ; Write protection off
cfg2&= _VCAPEN_OFF ; All VCAP pin functionality is disabled
cfg2&= _PLLEN_ON ; 4x PLL disabled
cfg2&= _STVREN_ON ; Stack Overflow or Underflow will cause a Reset
cfg2&= _BORV_19 ; Brown-out Reset Voltage (Vbor), low trip point selected.
cfg2&= _LVP_OFF ; High-voltage on MCLR/VPP must be used for programming
__CONFIG _CONFIG2, cfg2


#ENDCONFIG


#ENDIF


;----[Verify Configs have been specified for Selected Processor]----------------
; Note: Only include this routine once, after all #CONFIG blocks
#IFNDEF MCU_FOUND
#ERROR "No CONFIGs found for [" + __PROCESSOR__ +"]"
#ENDIF




ADCON1=%11110011 'enable adc internal reference and justify
FVRCON=%11001111 'set internal reference to 4.096V
'OSCCON=%01111000 'SET INTOSC TO 16MHZ
OSCCON=%11110000 'SET INTOSC TO 16MHZ
ANSELA=%00000000 'disable ADC on A
ANSELB=%00000000 'disable ADC on B
TRISC=%00000000 'set PORTC as output all
TRISB=%00000000 'set PORTB as output all
TRISA=%00000000 'set PORTA 2 as input, others as output
TRISE=%00000000 'set PORTE as output all
WPUB=%00000000 'DISABLE B PULL UPS
OPTION_REG=%100000000
LCDCON=%00000000 'disable LCD controller pins
LCDSE0=%00000000
LCDSE1=%00000000


;____[ For 12F/16F only - Interrupt Context save locations]_________________
wsave var byte $20 SYSTEM ' location for W if in bank0
'wsave var byte $70 SYSTEM ' Alternate save location for W
' if using $70, comment out wsave1-3
' --- IF any of these next three lines cause an error ?? -------------------
' Comment them out to fix the problem ----
' -- The chip being used determines which variables are needed -------------
wsave1 VAR BYTE $A0 SYSTEM ' location for W if in bank1
wsave2 VAR BYTE $120 SYSTEM ' location for W if in bank2
wsave3 VAR BYTE $1A0 SYSTEM ' location for W if in bank3
'---DO NOT change these-----------------------------------------------------
ssave VAR BYTE BANK0 SYSTEM ' location for STATUS register
psave VAR BYTE BANK0 SYSTEM ' location for PCLATH register


;----[ MIBAM Setup ]--------------------------------------------------------
BAM_COUNT CON 13 ; How many BAM Pins are used?
INCLUDE "MIBAM.pbp" ; Mirror Image BAM module


DEFINE OSC 32 'set oscillator speed


'DIGIT PINS
D0 VAR BYTE
D1 VAR BYTE
D2 VAR BYTE
D3 VAR BYTE
D4 VAR BYTE
D5 VAR BYTE
D6 VAR BYTE
D7 VAR BYTE
D8 VAR BYTE
D9 VAR BYTE
DF VAR BYTE 'TEMPERATURE SIGN
DT1 VAR BYTE 'DOTS
DT2 VAR BYTE


eval var byte 'eeprom reader
ELOC VAR BYTE 'EEPROM LOCATION


ASM
BAM_LIST macro ; Define PIN's to use for BAM


BAM_PIN (PORTB,3, D0)
BAM_PIN (PORTA,0, D1)
BAM_PIN (PORTB,4, D2)
BAM_PIN (PORTC,4, D3)
BAM_PIN (PORTB,5, D4)
BAM_PIN (PORTC,7, D5)
BAM_PIN (PORTB,2, D6)
BAM_PIN (PORTC,5, D7)
BAM_PIN (PORTB,1, D8)
BAM_PIN (PORTB,0, D9)
BAM_PIN (PORTC,6, DF)
BAM_PIN (PORTA,3, DT1)
BAM_PIN (PORTA,2, DT2)

endm
BAM_INIT BAM_LIST ; Initialize the Pins
ENDASM


D0=10: D1=10: D2=10: D3=0: D4=10: D5=10: D6=10: D7=10: D8=10: D9=10: DF=10 'set bam ports




mama:
high portc.3
pause 1
low portc.3
pause 1
goto mama


The above code should generate square wave output with 1ms high and 1ms low frequency. It does, but the "high" period slowly increases from 1ms to 1.3ms in several iterations of the loop, while "low" period remains the same. The jitter remains almost the same, and on values of PAUSE 20 and above, becomes almost non-existent.

Based on, that, I've developed a simple "protocol" which appears to be working, and is based on pulse with measurement:



receiver:
if portc.3=1 then
xcount=xcount+1 'increase counter while pin is high
else
gosub dcder 'go to decoder when it is low
xcount=0 'reset counter
endif
pauseus 1
goto receiver


dcder:
if xcount>10 and xcount<50 then rxvalue=0
if xcount>60 and xcount<100 then rxvalue=1
if xcount>110 and xcount<150 then rxvalue=2
if xcount>160 and xcount<200 then rxvalue=3
if xcount>210 and xcount<250 then rxvalue=4
if xcount>260 and xcount<300 then rxvalue=5
if xcount>310 and xcount<350 then rxvalue=6
if xcount>360 and xcount<400 then rxvalue=7


if rxvalue<>oldvalue then
gosub deco 'send to display
endif
oldvalue=rxvalue


return


This works quite reliably and no issues so far, but it is very slow and is only good as a temporal solution.
So I'm going to redesign PCB and use hardware serial ports on both transmitter and receiver, but will it 100% work? won't be it also affected by MIBAM?