Upgrading from older firmware¶

Firmware 109 changes how the driver measures and controls the motor more than any release before it. This page says what a driver keeps when it is updated to 109, what works at once, what has to be run, and how to go back. Updating through the bootloader is how to flash it.

The step from 109 to 110 is smaller: From 109 to 110.

From 109 to 110¶

Firmware 110 changes the gripper, adds #Estop and parameters 90 to 93, and saves brake or coast; nothing has to be measured again. A driver updated from 109 keeps everything it saved: the settings, the motor calibration, the encoder table and the anti-cogging. 110 keeps the idle behaviour in a slot of settings block 2 that was reserved.

Brake or coast (parameter 18) is saved from 110 on. A driver that has not saved it yet brakes in idle, as on 109; send #Coast and #Save once if it should coast.
The gripper calibration is kept in RAM, as it always was: calibrate the gripper after the update, as after every power-up. Positions 0 and 255 then lie a little inside the two ends (Gripper mode).
Motor GUI 0.7.0 is the version made for 110: its STOP sends #Estop over UART, and its presets write parameters 90 to 92 over CAN.

Going back to 109 is safe: 109 ignores the saved idle behaviour and brakes at every power-up again.

What is kept¶

Flashing never touches the EEPROM, where the configuration lives (Where it is kept). A driver updated from 106, 107 or 108 starts with everything it had:

the calibration: pole pairs, R, L, the angle offset, Kt, KV, the phase order and direction;
the CAN ID, the gains, the limits, the watchdog, the heartbeat, the thermistor and the gripper settings.

A driver from 101 to 105, saved before the configuration had checksums, is trusted and sealed at its first start of 109 when it was ever saved; 108 reset a board saved but never calibrated, CAN ID included, and 109 keeps it.

Release 103 and older from the main branch have no bootloader: such a board needs the programming adapter once (Installing the bootloader on an older board).

The new settings start from their defaults: tool id 0, the watchdog fallback 2000 ms / 0 mA / 1000 ticks/s / legacy, parameter 56 = 2.

What works at once¶

These need no measurement and are on from the first start of 109:

the direct PLL speed estimate at 60 Hz (The speed estimate);
commutation delay compensation, 166 µs (Commutation delay compensation);
the fast reading and PWM paths of the control loop;
the current-sense zero, measured at every start-up: a STEPFOC that read about 48 mA of current with nothing flowing reads under 1 mA;
undervoltage is not checked for the first 500 ms after power-up, so a supply that ramps up slowly no longer leaves a Vbus error behind (overvoltage always is);
the watchdog actions: the action byte the host already sends is acted on. Byte 0, what the Python library and the PAR6 runtime send, gives the legacy response, so an unchanged host sees no difference;
#Default and a settings reset keep the CAN ID;
the new CAN commands and parameters, the identity frame, telemetry, the capture, #Info's In force:, and sound through the motor.

What has to be run¶

These are off after the update, because nothing is saved for them yet; the encoder table and the anti-cogging cannot be switched on before they are measured:

the encoder table;
the anti-cogging;
the current-sense zero trim.

A calibration runs all three (What follows the calibration); so does #Encal all without a calibration.

What does not change by itself¶

The current gains. The saved Kpiq and Kiiq stay as they are: on a driver calibrated before 109 they are the old gains divided by 7, a loop of about 70 Hz (parameter 49 reads about 71). A calibration computes them again at 160 Hz; so does #Cbw 160 (or parameter 49 = 160), then #Save.
The velocity gains. The saved Kpv and Kiv stay as they are, typically 0.007 and 0.0005. Firmware 109 was tuned with 0.02 and 0.001, the PLL speed estimate allowing the higher gain; set them yourself (The velocity loop).

The whole upgrade¶

Flash 109 (Updating through the bootloader). #Info prints Software version: 109 and ends with In force:.
Check that the current limit suits the motor (#Ilim), then calibrate, with the shaft free: #Cal. The encoder table (at 80 % of the current limit), the anti-cogging and the trim follow, and everything is saved (about two minutes in all).
Set the velocity gains, for example #Kpv 0.02 and #Kiv 0.001 (a 21.5 mm stepper overshot 95 % at these on the bench; its presets use #Kpv 0.014 and #Kiv 0.00014), then #Save. (Calibration puts the current limit that was in force back when it ends; set #Ilim for your motor before step 2 if it is not.)

Check

#Info prints encoder table: on, anti-cogging: on (harmonics), saved.

With Motor GUI's presets: apply the preset first, then calibrate, then set the velocity gains. On firmware 109 a preset writes the current gains of a 160 Hz loop for its own R and L, the same a calibration computes from them, and its own velocity gains.

On a robot whose joints cannot turn freely, neither the calibration nor the runs after it can be made in place.

There skip step 2: the drive keeps the calibration it has, the new speed estimate, the delay compensation and the current-sense zero work at once, and #Cbw 160 with #Save gives the new current gains. The encoder table and the anti-cogging then stay off.

Going back¶

Flashing 108 or older over 109 is safe for the configuration: the calibration and the first settings block are unchanged, and older firmware ignores everything 109 saved from byte 256 on (tool id, fallback, anti-cogging, encoder table). Flash 109 again and they are all still there.

Not yet tested on the bench

The upgrade from 103 (programming adapter, then the bootloader, then 109) and a downgrade to 108 had not been tested on the bench when this page was written; the rest of what it says was, except where the pages it links say otherwise.

Comments

Found a mistake on this page? Edit it on GitHub · Report it · Get help · This page as Markdown · llms.txt · The whole manual as PDF