Memsic 2125 code?

[ScaleRobotics]

Hi Ed,

If I were you, I'd use the Parallax Memsic 2125 example, as it is probably the easiest of what I have seen. It uses a lookup table and simple math to approximate the angle, with pretty good accuracy within the 0 to 45 degree range (but error will increase as roll increases, since a lookup table can only take one axis into account at a time).

Parallax shows a table for angle/g from Memsic:


Straight from the nv92.pdf example:

Now, we could take the easy route and use 62.35 as our conversion factor, but this would give us
an error at low tilt angles that we really don't need to tolerate. What we'll do, then, is work
backward from this data and determine the output from the 2125 at the angles specified in the
chart. Then we can use LOOKDOWN to compare the 2125 output to the chart and a LOOKUP
table to determine the proper tilt conversion factor.
Calculating the maximum g-force output for a given range is done by dividing the arc for a range
by its conversion factor. This will give us the maximum g-force output for that range.

10 ÷ 57.50 = 0.17391
20 ÷ 58.16 = 0.34387
30 ÷ 59.05 = 0.50804
40 ÷ 60.47 = 0.66148
50 ÷ 62.35 = 0.80192

Remember that the g-force output from the BASIC Stamp code is in 0.001g units, so we'll
multiply the results above by 1000 for use in a LOOKDOWN table. Let's look at the code that
converts g-force to tilt.

Here is some of the code, modified only slightly for Picbasic:

xRaw    VAR Word ' pulse width from Memsic 2125
xGForce VAR Word ' x axis g force (1000ths)
xTilt   VAR Word ' x axis tilt (100ths)
idx     VAR byte ' table index
mult    VAR Word ' multiplier - whole part
frac    VAR Word ' multiplier - fractional part
HiPulse con 1    ' tells pulsin to look for highpulse, not lowpulse




Read_X_Force:
    PULSIN PORTB.1, HiPulse, xRaw
    xRaw = xRaw * 2
    ' g = ((t1 / 0.01) - 0.5) / 12.5% ' formula from data sheet
    xGForce = ((xRaw / 10) - 500) * 8 ' Modified for our PIC - Converts to 1/1000 g's
RETURN

Read_X_Tilt:
GOSUB Read_X_Force
    ' tilt = g x k
    '
    ' Select tilt conversion factor based on static
    ' G force. Table data derived from Memsic specs.
    LOOKDOWN2 ABS xGForce, <=[174, 344, 508, 661, 2000], idx
    ' above table are thousandths g results for 10,20,30,40 and max value degrees  
    LOOKUP idx, [57, 58, 59, 60, 62], mult          'integer lookup
    LOOKUP2 idx, [32768, 10486, 2621, 30802, 22938], frac 'gets fractional result
    ' G Force is divided by 10 to prevent roll-over errors at end
    ' of range. Tilt is returned in 100ths degrees.
    ' Parallax explains fractional lookup like this:
    ' 65,536 * .16 (this is from the fractional part of 20 degree, 58.16 number) = 10486
    xTilt = mult * (ABS xGForce / 10) + (frac ** (ABS xGForce / 10))
    Check_SignX:
    IF (xGForce.Bit15 = 0) THEN XT_Exit ' if positive, skip
    xTilt = -xTilt ' correct for g force sign
    XT_Exit:
RETURN

[Ramius]

Thanks Walter! You are the best! Ed

[Ramius]

Hi Walter and everyone!
Okay so I corrected the code in the Read_X_Force as the "return" caused the program to stop and the formula said to divide by 8 not multiply. With the digital scope I was able to know the exact timing of the pulses. The catch is that the devide puts out a pulse reading from 1810 to 3132 so the formula does not work. I attached word files to show the code, what the pulse measurement is at level, what the code does with level and what the code does with the device at +90 degrees (up) and -90 degrees (down). The original code uses 10 ms and the scope says it is 9.9 ms so rather than x2 I made it x20 to divide by 99. The problem with the code seems to be with the 50% duty cycle value? Please let me know what you think. Thanks, Ed


Pulse Widths.doc
Tilt Code.doc
UP.doc
Level.doc
Down.doc

[ScaleRobotics]

Okay so I corrected the code in the Read_X_Force as the "return" caused the program to stop and the formula said to divide by 8 not multiply.

Ed, are you refering to the place in the formula where there was a /12.5% ? If so, this is simplified by *8.

See this part of the previously attached document:
"Since our T1 time is in microseconds and there are 1000 microseconds in a millisecond, we need
to multiply the T2 value and 0.5 by 1000 to adjust the equation. This makes the math easy for the
Stamp and gives us an output with a resolution of 0.001g. Finally, the divide by 12.5% part of the
equation is converted to multiplying by eight ( 1 / 0.125 = 8 ) – we couldn't ask for a value much
more convenient than that."