I’m using ODrive (fw-v0.5.4) with hoverboard motors and AMT103 encoders, and it all works quite well now, but it worries me that I get highly inconsistent phase_offset_float values after ENCODER_OFFSET_CALIBRATION.
Here are the commands that I’ve run in odrivetool to demonstrate the issue:
odrv0.erase_configuration()
odrv0.config.enable_brake_resistor = True
odrv0.config.brake_resistance = 2
odrv0.config.gpio12_mode = GPIO_MODE_ENC1
odrv0.config.gpio13_mode = GPIO_MODE_ENC1
odrv0.config.gpio14_mode = GPIO_MODE_DIGITAL
odrv0.axis1.motor.config.pole_pairs = 15
odrv0.axis1.motor.config.resistance_calib_max_voltage = 4
odrv0.axis1.motor.config.requested_current_range = 25
odrv0.axis1.motor.config.current_control_bandwidth = 100
odrv0.axis1.motor.config.torque_constant = 8.27 / 16
odrv0.axis1.encoder.config.mode = ENCODER_MODE_INCREMENTAL
odrv0.axis1.encoder.config.cpr = 8192
odrv0.axis1.controller.config.control_mode = CONTROL_MODE_VELOCITY_CONTROL
odrv0.save_configuration()
odrv0.axis1.requested_state = AXIS_STATE_MOTOR_CALIBRATION
print(odrv0.axis1.motor.error, odrv0.axis1.motor.config.phase_inductance, odrv0.axis1.motor.config.phase_resistance)
0 0.0004988669534213841 0.24248993396759033
odrv0.axis1.motor.config.pre_calibrated = True
import time
for d in range(5, 201, 5):
print("%3.0f:" % d, end="")
odrv0.axis1.encoder.config.calib_scan_distance = d
for repeat in [1, 2, 3]:
odrv0.axis1.encoder.config.phase_offset_float = 0
for state in [AXIS_STATE_ENCODER_INDEX_SEARCH, AXIS_STATE_ENCODER_OFFSET_CALIBRATION]:
odrv0.axis1.requested_state = state
time.sleep(0.2)
while odrv0.axis1.current_state != AXIS_STATE_IDLE:
time.sleep(0.1)
time.sleep(0.2)
print(" %6.3f" % odrv0.axis1.encoder.config.phase_offset_float, end="")
print()
(An ENCODER_INDEX_CALIBRATION before each offset calibration may or may not be necessary;
I just wanted to make sure that all offset calibrations start at the same motor/encoder position, eliminating the possibility that these inconsistent values are caused by the random starting position.)
And the resulting phase_offset_float values for various calib_scan_distances:
5: 1.117 -0.093 -0.341
10: 0.418 -0.335 0.499
15: 0.420 -0.203 0.077
35: 0.289 -0.447 0.231
40: -0.129 -0.267 -0.315
45: -0.216 -0.197 -0.058
50: 0.385 0.237 0.345
55: -0.397 -0.449 -0.150
60: 0.311 0.145 -0.482
65: 0.286 0.138 -0.226
70: -0.293 -0.342 0.312
75: 0.038 0.090 0.161
80: 0.373 0.256 0.017
85: 0.443 -0.412 -0.398
90: 0.160 0.196 0.190
95: 0.250 0.248 0.333
100: 0.304 -0.473 0.453
105: 0.099 0.045 0.169
110: 0.207 0.321 0.245
115: 0.324 0.314 0.346
120: 0.304 -0.163 0.129
125: 0.239 0.337 0.337
130: -0.492 -0.397 -0.393
135: -0.024 0.017 -0.008
140: -0.089 -0.226 -0.210
145: 0.285 0.429 -0.291
150: -0.017 -0.038 -0.100
155: -0.278 -0.223 -0.164
160: 0.172 0.281 0.329
The first column shows calib_scan_distance while the other three show the resulting phase_offset_float. As you can see the value of phase_offset_float doesn’t seem to converge with increasing calib_scan_distance values, and, worse, repeated offset calibrations with the same distance tend to vary considerably.
Is this normal? Or at least is it something that I should not be worried about?
FWIW, when I used the same motors with Hall effect sensors, then phase_offset_float was always quite close to 0.5 % 1 as mentioned in docs/hoverboard.md.