HEDS5540 optical encoder

[louislouis]

For testing and learning the 16F628A is seems to be OK. But now I try to convert the pulse reading to distance. One full encoder turn generates 1440 pulses. The wheel diameter is 25mm, the circumference is 157mm. I want reading precision to 0,1mm. I try basic math like:

Main_Loop: 
if Flag = 1 then 
let counter = (enc_counter/36) * 157 / 4
Lcdout $fe, 1                    
lcdout  Dec counter     ; display counter value on LCD
Flag = 0
endif
goto Main_Loop

But the precision is terrible, can you explain me the right way?

[richard]

But the precision is terrible, can you explain me the right way?

157/1440 = .109 mm per encoder tick

.109 * 256 * 10 = 279

distance * 10 = (counts * 279)/256

to use pbp's higest resolution integer functions

counter = enc_counter */ 279 

lcdout  Dec counter/10 ,".", dec counter//10     ; display counter value on LCD

problem here is that when enc_counter > 60000 [or thereabouts] the internal 32bit result will overflow and the result will be wrong
you could look at nbit math


ps you could also use div32

counter = enc_counter * 1000
counter =  div32 917

lcdout  Dec counter/10 ,".", dec counter//10     ; display counter value on LCD

[louislouis]

Thank You Richard for helping me. That's great, now the "core" of my distance meter works perfectly. Yes, the max. distance what I can measure is near six and a half meter which is enough for me. If I want to measure more than 6,5m simply zeroed the reading and start count from that point. Now I must implement negative reading when I go to opposite direction from zero the reading must be from -1..... to -655,3.
Your explanation helping me a lot, once again Thank You.

[louislouis]

I finally assembled my distance meter based on rotary optical encoder and precision wheel. But still I have problem with math and precision.
For example, wheel circumference is 157mm, and when I take a 30 revs the result must be 4710mm, but the MCU calculates only 4708,1mm which is 1,9mm difference. Is there a way to improve the precision?
I use Richard's math:
157/1440 = .109 mm per encoder tick
.109 * 256 * 10 = 279
distance * 10 = (counts * 279)/256

counter = enc_counter */ 279 
lcdout  Dec counter/10 ,".", dec counter//10     ; display counter value on LCD

The precision 0,1mm is enough for me, the max. measurable distance want be 6000mm. I try some change, if I improve the precision to 0.01mm then the max. measurable distance shortened to 600mm and this not enough.

[richard]

you could look at nbit math or use a pic18 and long vars

you can glean a bit more precision breaking the calc up a bit

where 43200 counts would give 4709.2

 
frac = enc_counter.lowbyte * 1090         
posn1  = div32 100          ;mm*100          ;mv 2790
posn2 = (enc_counter.highbyte&15) * 2791  ; mm*100   mv 41865
frac = (enc_counter.highbyte>>4) * 44657
posn3 =  div32 100                       ;mm     mv6698

posn3=  posn3 + (posn1+posn2)/100       ;mm
frac=   ((posn1+posn2)//100 +5)/10        ;round it up
serout2 PORTa.0,84, [13,10,#enc_counter,9,dec posn3,".",#frac       ]

[louislouis]

Hello Richard, thank you again for help. Your code works and take a little improvement. I make some changes, first got a smaller precision wheel dia. 25mm (circ. 78,53975mm) and divide the encoder reading /2. Now the encoder reading is only 720 per one rev.
The code is as follows:

Main_Loop:
enc_counter = enc_counter /2
if Flag = 1 then 
frac = enc_counter.lowbyte * 1090         
posn1  = div32 100          ;mm*100          
posn2 = (enc_counter.highbyte&15) * 2794  ; mm*100   
frac = (enc_counter.highbyte>>4) * 44680
posn3 =  div32 100                       ;mm     
posn3=  posn3 + (posn1+posn2)/100       ;mm
frac=   ((posn1+posn2)//100 +10)/10        ;round it up
 
Lcdout $fe, 1
lcdout  "Dist: ",dec posn3,".",#frac
Flag = 0
endif 
goto Main_Loop

Now I can measure up to 7100mm, but still more or less precise.
Next step will be order one of the PIC18x mcu and try long variables.
Here are suitable to use the same encoder reading code (written in ASM) and simple change the word var to long var?
I think its not that easy. When I have the 18x PIC I think, I will need more help with this project. Thanks.

[richard]

frac= ((posn1+posn2)//100 +10)/10 ;round it up

is incorrect , it will round the number up always

frac= ((posn1+posn2)//100 +5)/10 ;round it up

is the way to round the number up if the last digit is 5 or more