No luck with auto correlation curves using Calibrate_All

Martin_H

I'm running the Calibrate_All with a home made BS2E robot and wheel encoders. I then used the Calibrate All Windows program to download and graph the data which you can see below.

The pulse widths versus velocity curves look great (except for a slight flattening around the origin for the left wheel). To me that says that both encoders are working, and that's been my experience with some other test programs I've run. But what is strange is the photo sensor and auto correlation curves. The curves don't look half bad, except the red and green curves are 180 degrees out of phase. I guess I'm wondering what's up with that? I read the source code to "Calibrate_All.bs2" and it samples the photocell once, wires a word, and then pulses both wheels. So I don't understand how this could happen. Here's a snippet of the code that does this.

  CASE 3

    PAUSE 5000
    FOR i = 0 TO 512 STEP 2
      HIGH Photo
      PAUSE 6
      RCTIME Photo, 1, Value
      WRITE i + 64, Value.HIGHBYTE
      WRITE i + 65, Value.LOWBYTE
      Dist(LEFT) = i.BIT1
      Dist(RIGHT) = 1 - Dist(RIGHT)
      Dir(LEFT) = FWD
      Dir(RIGHT) = BAK
      Veloc = 1
      GOSUB DoMove
    NEXT
    END

The one thing that struck me as odd was the line Dist(RIGHT) = 1 - Dist(RIGHT). I guess I would have expected Dist(RIGHT) = 1 - Dist(LEFT).

Update: I made that change and it made no different. The data is still 180 degrees out of phase. Here's the raw correlation data.

Copy and paste these DATA statements into your BASIC Stamp programs:

DATA @0, 6,15,23,28,33,38,43,47
DATA     52,57,62,68,74,80,86,92
DATA     10,17,21,28,34,39,44,50
DATA     55,61,66,72,78,84,89,95
DATA     16,19,23,28,32,37,41,46
DATA     51,55,60,66,72,77,83,89
DATA     18,20,25,31,36,40,45,51
DATA     57,62,68,75,81,87,93,100

Autocorrelation data for Calibrate_All.exe:

8,10,11,18,23,55,64,83
97,117,138,155,163,172,190,213
227,242,228,223,228,245,264,287
298,333,399,430,442,402,351,286
229,188,173,156,148,129,116,98
84,30,9,8,8,7,7,9
12,19,36,59,71,89,104,118
125,148,178,200,211,202,190,191
194,201,214,227,239,267,283,306
322,319,291,277,262,237,206,166
138,116,98,83,70,54,12,5
5,5,5,5,7,10,17,39
57,66,83,95,112,134,163,181
188,191,193,187,185,184,185,186
187,190,195,203,215,224,258,293
305,302,270,235,211,171,159,128
103,89,76,67,40,9,5,5
4,4,5,8,11,23,43,60
74,87,100,115,140,171,189,193
186,179,178,180,184,194,207,216
238,264,294,300,272,240,189,160
128,113,90,77,68,49,12,5
5,5,4,4,6,9,15,37
57,68,84,96,111,133,162,173
181,185,180,172,174,180,189,201
218,233,276,304,315,291,263,215
186,169,160,144,127,103,88,76
64,37,6,5,5,5,5,5
9,13,27,48,65,77,88,101
119,139,168,180,186,177,176,179
185,195,208,221,232,248,294,306
310,312,305,280,247,199,163,135

END

Martin_H

I reran the fourth calibration program again and made a great deal of effort to improve the lighting. I added a little hood around the CDS cell, now it only sees light from the front. I took it into a room with one sunny window, two interior walls, and one window away from the sun. I made sure the robot wasn't in direct sunlight, but was exposed to the sunny window. This got rid of the spike, valley, spike in the brightest part of the room, and the curves don't look bad to me. But the Callibrate_all.exe program still doesn't like them.

I'm not sure what I can do to improve my data.

Copy and paste these DATA statements into your BASIC Stamp programs:

DATA @0, 6,15,23,28,33,38,43,47
DATA     52,57,62,68,74,80,86,92
DATA     10,17,21,28,34,39,44,50
DATA     55,61,66,72,78,84,89,95
DATA     16,19,23,28,32,37,41,46
DATA     51,55,60,66,72,77,83,89
DATA     18,20,25,31,36,40,45,51
DATA     57,62,68,75,81,87,93,100

Autocorrelation data for Calibrate_All.exe:

88,82,76,83,92,96,99,102
122,133,178,212,240,279,511,772
844,835,922,951,958,960,960,975
981,980,1006,1099,1134,1148,1165,1214
1357,1435,1637,1744,1837,1808,1668,1451
1277,1111,1016,944,819,609,328,185
130,106,90,79,78,86,90,91
95,102,109,115,129,150,191,212
240,329,416,689,783,844,852,848
869,929,961,961,966,972,972,992
1103,1118,1159,1283,1509,1697,1815,1800
1677,1470,1300,1135,1041,946,853,703
388,213,141,110,94,83,78,86
90,92,93,99,107,117,139,160
181,192,282,407,765,817,840,856
840,923,961,961,972,970,1026,1118
1147,1249,1323,1491,1634,1764,1780,1673
1505,1309,1148,1041,950,849,734,512
354,239,202,161,135,116,107,95
82,77,75,79,87,92,107,126
156,184,200,324,464,742,802,820
860,946,957,958,958,970,979,990
1004,1048,1112,1164,1179,1211,1335,1509
1693,1789,1803,1776,1646,1560,1402,1248
1136,1070,1001,951,863,798,556,408
220,177,125,111,104,96,86,80
78,75,77,80,84,90,97,103
112,117,127,164,223,278,340,575
792,837,842,926,950,963,961,972
991,1005,1086,1153,1203,1337,1492,1600
1770,1833,1861,1809,1690,1579,1406,1292
1166,1097,1073,1047,989,941,908,852

END

Phil Pilgrim (PhiPi)

Martin,

What the program appears to be complaining about is the increasing period, not the shape of the curve. Are you doing the calibration on carpet by any chance? A hard surface works best.

-Phil

Martin_H

Thanks for the response. The first try was on a stone counter, the second was on a card table. Let me retry with a few tweaks and see if I can fix that.