Data processing in array

[Ioannis]

There is an array of n elements. Each element stored the pulse width from a PulsIn. Then according to pulse width, the bit_array bit will be 0 or 1. But if it is out of range then a new data set will be collected.

If the byte_array(x) is between a and b then bit_array.0(x)=0

If the byte_array(x) is between c and d then bit_array.0(x)=1

if complete n elements then continue, else goto parse new byte_array data.

Seems easy but could not find a way to exit the loop!

Any help please?

parse byte_array

find_bit:  
            if byte_array[index]>145 then    '>725ìs
                if byte_array[index]<180 then
                    bit_array.0[index]=0
                    index=index+1
                    if index<n then find_bit
                endif
            endif

            if byte_array[index]<100 then    '>725ìs
                if byte_array[index]>75 then
                    bit_array.0[index]=1
                    index=index+1
                    if index<n then find_bit
                endif
            endif

Ioannis

[richard]

i would do it this way

 sample_data var byte[32] 
fail var bit
 sample var byte
 a_one  var byte
 a_zero var byte
 result var byte[4]
 tolerance var byte
 
    tolerance = 20
    a_one = 160  ; mean value of a one bit
    a_zero = 90   ; mean value of a zero bit
 
 
clear_result:
    result[0] = 0
    result[1] = 0
    result[2] = 0
    result[3] = 0
get_result: 
    sample = 32
    fail = 0
    while sample && ! fail
        sample = sample-1
        if   (abs(sample_data[sample] - a_one)< tolerance ) then 
           result.0[sample] = 1
        else 
            if  (abs(sample_data[sample] - a_zero)> tolerance ) then fail = 1
        endif 
    wend
    if fail then   
        result bad
    else
        break out the champagne
    endif

[Ioannis]

Thanks Richard.

Makes sense though kinda heavy in respect of calculations.

Ioannis

[richard]

kinda heavy in respect of calculations.

maybe but it has positive indication of failure , it ceases testing on failure and it actually detects failures.
i would not be quibbling over a few words of code space when the entire process of encoding the bit stream into an unnecessary byte array
of times instead of directly into a bit array on detection would be heaps more efficient.


if the correct conditions can be met the pbp abs command can substituted for byte results like below


failure detection comparison

'parse byte_array define OSC 8
 bit_array var byte[4]
 byte_array var byte[32]
 arraywrite  byte_array,[80,80,160,80,80,160,80,160,_
                         80,160,80,80,160,80,80,160,_
                         160,80,80,80,80,160,160,80,_
                         80,80,80,160,80,110,160,160]
 n           var byte 
 index      var byte
 trisa.1 =0
 porta=2
 pause 1000
 serout2 porta.1 ,84 ,["ready",13]
 
 n = 32 
 index = 0


find_bit:  
            if byte_array[index]>145 then    '>725ìs
                if byte_array[index]<180 then
                    bit_array.0[index]=0
                    index=index+1
                    if index<n then find_bit
                endif
            endif


            if byte_array[index]<100 then    '>725ìs
                if byte_array[index]>75 then
                    bit_array.0[index]=1
                    index=index+1
                    if index<n then find_bit
                endif
            endif
 serout2 porta.1 ,84 ,[bin8 bit_array[3], bin8 bit_array[2], _
                         bin8 bit_array[1], bin8 bit_array[0],13]    
       
 ;my way           
    tolerance con 20
    a_one con 90  ; mean value of a one bit
    a_zero con 160   ; mean value of a zero bit
    fail var bit


clear_result:
    bit_array[0] = 0
    bit_array[1] = 0
    bit_array[2] = 0
    bit_array[3] = 0
get_result: 
    n = 32
    fail = 0
    while n && (!fail)
        n = n-1
        index = byte_array[n] - a_one;    bytewise abs shortcut
        if index.7 then index = ~index+1;    bytewise abs shortcut
        if   index <tolerance  then  ;     (abs(byte_array[n] - a_one))  < tolerance 
           bit_array.0[n] = 1
        else 
         index = byte_array[n] - a_zero ;    bytewise abs shortcut
        if index.7 then index = ~index+1 ;    bytewise abs shortcut
            if  index   then fail = 1   ;(abs(byte_array[n] - a_zero))>tolerance
        endif 
        serout2 porta.1 ,84 ,["."]
    wend
    if fail then 
       serout2 porta.1 ,84 ,[13,"fail",13]  
    else 
        serout2 porta.1 ,84 ,[13,bin8 bit_array[3], bin8 bit_array[2], _
                         bin8 bit_array[1], bin8 bit_array[0],13]    
    endif

[richard]

result of sim

[Ioannis]

...when the entire process of encoding the bit stream into an unnecessary byte array
of times instead of directly into a bit array on detection would be heaps more efficient...

Sure I agree but could not find a fast way to encode directly the PWM bit stream into bits. The bits are coming from an RF radio receiver and have the typical 66-bit format of a keeloq transmission: 16ms guard time (low), 12 bits of 400us/400us as preamble, a 3-4ms low header and then the 66 bits of the data.

My routine to receive all these bits is based on the PulsIn command in a while-wend loop:

code        VAR byte[9] BANK0   ' PWM read in
code1       var byte[66]            ' Keeloq bits to be decrypted

KEY_0       var byte BANK0
KEY_1       var byte BANK0
KEY_2       VAR BYTE BANK0
KEY_3       VAR BYTE BANK0
KEY_4       VAR BYTE BANK0
KEY_5       VAR BYTE BANK0
KEY_6       VAR BYTE BANK0
KEY_7       VAR BYTE BANK0
mask        VAR BYTE BANK0
CSR_0       VAR BYTE BANK0
CSR_1       VAR BYTE BANK0
CSR_2       VAR BYTE BANK0
CSR_3       VAR BYTE BANK0
pass        VAR BYTE BANK0
CNT1        VAR BYTE BANK0
CNT2        VAR BYTE BANK0

temp        var word

OVER:
while 1
main:
index=0
guard:
    pulsin portb.0,0,temp     'find header of 16ms
    if temp<2500 then main   ' PIC is clocked at 8MHz, so a PulsIn step is 5usec.
    if temp>4400 then main   ' 5usec x 2500= 12.5ms up to 5usec x 4400=22ms
preamp:
    pulsin portb.0,1,temp     'then skip 12 pulses of 50% Duty, 400usec
    if temp<50 then main      
    if temp>90 then main      
    index=index+1
    if index<11 then preamp
         
    index=0 
pulsetrain:                   'now collect 66 pulses ~800/400 usec
    pulsin portb.0,1,code1[index]
    index=index+1
    if index<66 then pulsetrain

    index=0
find_code:  'convert pulse width to binary 66 bit stream
            if code1[index]>130 then
                if code1[index]<170 then
                    code.0[index]=0
                endif
            endif
            
            if code1[index]<80 then
                if code1[index]>55 then
                    code.0[index]=1
                endif
            endif
            
            index=index+1
            if index<66 then find_code
            
            'Display the encrypted and decrypted array
            hserout ["Rec: ",bin2 code[8]," ",bin8 code[7]," ",bin8 code[6]," ",bin8 code[5]," ",bin8 code[4],":",bin8 code[3]," ",bin8 code[2]," ",hex2 code[1]," ",hex2 code[0],13,10]  'display pulse train
            CSR_0=CODE[0]:CSR_1=CODE[1]:CSR_2=CODE[2]:CSR_3=CODE[3]
            GOSUB DECRYPT_K
            hserout ["Dec: ",bin2 code[8]," ",bin8 code[7]," ",bin8 code[6]," ",bin8 code[5]," ",bin8 code[4],":",bin8 CSR_3," ",bin8 CSR_2," ",hex2 CSR_1," ",hex2 CSR_0,13,10,13,10]
            
wend

Obviously, cannot show the DECRYPT_K routine... Sorry.

The above code is not near your fail safe program and will try to fit in the code space after a clean up of the rest code.

Ioannis

[Ioannis]

The more I observe your code and way of thinking, the more I see its briliance.

Thank you again Richard,
Ioannis

[richard]

a couple of ways

pulsetrain:  
index=67 
while index
while  !portb.0 :wend   ;wait for hi
index=index-1
pauseus 600                  ;bit period=3*te ie 3*400=1200uS
code.0[index]= ~ portb.0      ;sample @midpoint of bit period
while  portb.0 :wend    ;wait for lo
wend

;or with some simple time supervision

fail var bit
timer var word ext
@timer = TMR1L
PRELOAD CON 65535-820  ;@8MHz  timer prediv 2  each tick=1uS




pulsetrain:  
index=67 
t1con=0
fail=0


while index
timer=preload
while  !portb.0 :wend   ;wait for hi
t1con=$21
index=index-1
pauseus 600                  ;bit period=3*te ie 3*400=1200uS
code.0[index]= ~ portb.0      ;sample @midpoint of bit period
while  portb.0 && !pir1.0 :wend    ;wait for lo
t1con=0
fail=fail | pir1.0
wend

or sample @50,250,450,650,850 uS past transition to hi
where valid result is
11000 for a 1
11110 for a 0

there are endless methods

[Ioannis]

My simple thinking lead me to this version ditching the big array:

    index=0 
pulsetrain:                   'now collect 66 pulses ~800/400 ìsec
    pulsin portb.0,1,temp1
    if temp1<55 then main
    if temp1>170 then main
    if temp1<130 then
        code.0[index]=1
    else
        code.0[index]=0
    endif
    index=index+1
    if index<66 then pulsetrain

And then came your masterpiece of coding with TMR1...! Amazing!

It just needs to be reversed (the code array) I guess. No big deal.

Thanks,
Ioannis

[richard]

In a noisy environment you can over sample a bit to be sure of data validity, i expect the serial data from a keeloq
chip to be quite clean so a single sample point would suffice

sample @50,250,450,650,850 uS past transition to hi
where valid result is
11000 for a 1
11110 for a 0

sample var bitfail var bit


fail=0
pulsetrain:
index=67 
while index && !fail
while  !portb.0 :wend   
pauseus 50               ;50
if   !portb.0 then  fail = fail | 1
pauseus 200              ;250
if   !portb.0 then  fail = fail | 1
pauseus 200              ;450
sample=portb.0
pauseus 200               ;650
if   portb.0^sample then  fail = fail | 1
pauseus 200               ;850
if   portb.0 then  fail = fail | 1
index=index-1

code.0[index]= !sample

wend

[Ioannis]

In general the data stream on a good signal level is very clean. But when there is no RF signal, the radio module will produce a lot of noise that is converted to random pulses through the data slicer. This creates a lot of bad data and false triggers on a non rolling code systems if digital filtering is not applied. With the keeloq encryption it is safe to assume that such triggers will not happen. The probability to have 66 bits in a correct timing and also in the correct sequence, not mentioning the other safety measures of keeloq like counter etc, is really low if not zero.

When you referred to 5 sampling points I was thinking of timers etc to maintain a fixed sampling point in time...

Your code showed the application of KISS in a fine way!

Thank you for your great ideas,
Ioannis

[richard]

The probability to have 66 bits in a correct timing and also in the correct sequence, not mentioning the other safety measures of keeloq like counter etc, is really low if not zero.

i would not expect to ever receive a malformed bit from a keeloq serial stream , the data may be useless however.
single point sampling should be ok imho

[Ioannis]

Agree.

Though I have checked various implementations of RF remote controls and many have wide tolerances from the specified timings.

I am very much sure that those chips are bad copies of the original Microchip one. But are out there... If interested, I have many test of original and copy chips with the timings in excel file.

Ioannis