Problems with decimal values and displaying them. Please help!

RussM · 2013-02-27 08:02

I am new to SPIN so I am sorry if this problem seems trivial to you all. I am working a project that using a rotary encoder to measure distance in thousands of an inch. What I want to do is each time the encoder value gets incremented, to add 0.000983 to my overall distance. When the encoder value gets decremented, i want to subtract 0.000983 from my overall distance. My problem is that I can't represent decimal values at all. I tried just writing the value 0.000983 to the screen and it printed out 981522434 instead. I tried utilizing the f32 object but to no avail. Any help would be greatly appreciated!

CON
  _clkmode = xtal1 + pll16x
  _xinfreq = 5_000_000
  '_xinfreq = 6_250_000
  
  TX_PIN = 0
  BAUD = 19_200
  'MS_001 = CLK_FREQ / 1_000
  'CLK_FREQ = ((_clkmode-xtal1)>>6)*_xinfreq
  
  LCD_PIN       = 27
  LCD_BAUD      = 19_200
  LCD_LINES     = 4
  LCD_COLS      = 20


  ENCODER_PIN   = 4


OBJ
  quad  :       "QuadDecoder"
  LCD    :       "FullDuplexSerial"
  f32           : "F32"
  f32_orig      : "Float32Full"
  fs            : "FloatString"


VAR
  long  offset                                          ' example variable that will be accumulated to
  long distance
  long temp
  long distance_tick
  long f1
  long fA
  


PUB main


  LCD.start(TX_PIN, TX_PIN, 00, 19_200)    'Initialize FullDuplexSerial.spin
  f32.start
  f32_orig.start 


  offset := 0                                          ' initialize the accumulator
                                                        ' You can set it to any desired value
  LCD.tx($0C)
  quad.start(ENCODER_PIN, @offset)                      ' start the encoder reader




    LCD.str(string("Value = "))
    LCD.tx($0D)
    LCD.str(string("Distance = "))
    
    distance_tick := 0.000983
    distance := 0
    temp := 25


    repeat
      if offset > temp
         distance := distance + distance_tick
      if offset < temp
        distance := distance - distance_tick
      LCD.tx($89)
      LCD.dec(offset)       'Encoder always initializes at 25 for some reason
      LCD.tx(32)
      LCD.tx($9F)
      LCD.dec(distance_tick)
      temp := distance
      waitcnt(clkfreq/10 + cnt)

kwinn · 2013-02-27 09:06

The problem with using floating point for this is rounding errors, and they accumulate when you do the calculations like you are doing them. It would be more accurate and simpler to use encoder counts as the basis for your positioning, convert the encoder count to position, and convert desired position to the encoder count.

Mike Green · 2013-02-27 09:21

In addition to kwinn's comments, the .dec method works with integers, not floating point. There is a specific floating point input/output conversion object called FloatString that you'd need to use if you still want to use floating point. It converts floating point values to strings in a temporary buffer suitable for display with the .str method.

You initialize some variables like distance to zero. It just so happens that a floating point zero is the same as the integer zero. It would be better from a documentation standpoint if you used 0.0 when you want to initialize a variable to zero for floating point use. There's a similar issue with temp and offset. You either need to convert distance to an integer using .FTrunc or .FRound to assign it to temp or you need to treat offset and temp as floating point throughout your program. That includes calling the floating point compare routine to compare the two values in your if statements.

Duane Degn · 2013-02-27 10:11

I agree with Kwinn's suggestion to use encoder pulses as you base unit.

If you don't want to use encoder pulses as the base unit, you use millionths of an inch as the unit instead of inches. Then your math could still be done with integers.

If using kwinn's method, you could do the conversion when needed (for display purposes) using millionths of inches in the calculation and then use the method "DecPoint" to place the decimal point in the desired position.

Here's the DecPoint method.

PUB DecPoint(value, denominator)
  if value < 0
    Pst.Char("-")
    -value
      
  if value => denominator
    result := value / denominator
    Pst.Dec(result)
    value //= denominator     
  else    
    Pst.Char("0")
  Pst.Char(".")  
  repeat while denominator > 1
    denominator /= 10
    if value => denominator
      result := value / denominator
      Pst.Dec(result)
      value //= denominator
    else
      Pst.Char("0")

An example of how to use it can be found in post #4 of this thread.

RussM · 2013-02-27 10:33

Thank you both Mike Green and kwinn for your help! I managed to get my code working using the FloatString object as follows:

CON
  _clkmode = xtal1 + pll16x
  _xinfreq = 5_000_000
  '_xinfreq = 6_250_000
  
  TX_PIN = 0
  BAUD = 19_200
  'MS_001 = CLK_FREQ / 1_000
  'CLK_FREQ = ((_clkmode-xtal1)>>6)*_xinfreq
  
  LCD_PIN       = 27
  LCD_BAUD      = 19_200
  LCD_LINES     = 4
  LCD_COLS      = 20


  ENCODER_PIN   = 4


OBJ
  quad  :       "QuadDecoder"
  LCD    :       "FullDuplexSerial"
  f32           : "F32"
  f32_orig      : "Float32Full"
  fs            : "FloatString"


VAR


  long  offset                                          ' example variable that will be accumulated to
  long distance
  long temp
  long distance_tick
  long f1
  long fA
  


PUB main


  LCD.start(TX_PIN, TX_PIN, %1000, 19_200)    'Initialize FullDuplexSerial.spin
  f32.start
  f32_orig.start 


  offset := 0                                          ' initialize the accumulator
                                                        ' You can set it to any desired value
  LCD.tx($0C)
  LCD.tx($11)
  quad.start(ENCODER_PIN, @offset)                      ' start the encoder reader




    LCD.str(string("Value = "))
    LCD.tx($0D)
    LCD.str(string("Distance = "))
    
    distance_tick := 0.000983
    distance := 0.0
    temp := 25.0


    repeat
      if offset > temp
         distance := f32_orig.FAdd(distance,distance_tick)
         temp := offset
      if offset < temp
        distance := f32_orig.FAdd(distance,distance_tick)
        temp := offset
      LCD.tx($89)
      LCD.dec(offset)       'Encoder always initializes at 25 for some reason
      LCD.tx(32)
      LCD.tx($9F)
      LCD.str(fs.FloatToFormat(distance, 8, 6))
      waitcnt(clkfreq/10 + cnt)

Mike Green · 2013-02-27 10:40

It's still not going to work properly because the comparisons in the if statements are done with integer arithmetic, but you're initializing temp to 25.0 which is not the same as 25. You probably want to initialize temp to 25.

RussM · 2013-02-27 10:47

I changed temp to 25 when I initialize it. Also a new problem. I tried to change FAdd to FSub in the second if statement and now it wont display any information at all anymore.

MagIO2 · 2013-02-28 23:43

I also do not understand why you want to stay with float numbers. Fix-point integer is much easier, faster, needs no additional code and no additional COG for the float AND it is more accurate!
( http://en.wikipedia.org/wiki/Floating_point#Accuracy_problems )

An 32 bit integer can have the values ranging from 0 to 4,294,967,295 (unsigned). So, in your case you can use the least significant digits for the fraction and still have a range of 0-4,294.

distance_tick:=0_000983
distance:=0_000000
offset:=25_000000
...
if step_direction
  distance += distance_tick
else
  distance -= distance_tick
...

(I don't really understand how the offset and temp variables should work in your code. The offset is changed by the encoder and the code probably works fine when the encoder is turned slow and/or the code to calculate the new distance is fast enough. But in any other case the offset might have changed by a value that is equivalent to more than just one single distance_tick!)

The only thing you have to adapt is the dec-method which needs to insert a "." at the fixed position and start printing the number at least with 7 zeros (0.000000)!

If the range is not enough it's easy to add another long and do keep the fractional part in one long, the integer part in another.

I would calculate the distance in the encoder-code directly! The encoder code already has a loop to read the encoder and increment/decrement an offset. So, why add another loop (which wastes a COG) to convert the encoder ticks into a distance?

Duane Degn · 2013-03-01 00:18

I agree with MagIO2 about using integer math.

MagIO2 wrote: »

The only thing you have to adapt is the dec-method which needs to insert a "." at the fixed position and start printing the number at least with 7 zeros (0.000000)!

The "DecPoint" method I posted above does this for you.

Problems with decimal values and displaying them. Please help!

Comments