GRBL Travel Speed

Stoli · March 17, 2025, 5:30pm

Does anyone know the formula(s) that FluidNC uses to determine motor speeds when approaching and departing from gcode vertices? My time estimates for long plots can be way off and I’m trying to come up with a formula that accounts for the deceleration and acceleration, which seem to be very significant for certain plots where the tool head makes a lot of large (90-180 degree) directional changes between points.

jeyeager · March 17, 2025, 6:41pm

I haven’t tried to digest this yet but starting with the comment on line 58. It seems to calculate start and end speeds of each move using the acceleration and re-evaluate every move when a new move is added. This seems like a hard problem to solve.

github.com/bdring/FluidNC

FluidNC/src/Planner.cpp

main

// Copyright (c) 2011-2016 Sungeun K. Jeon for Gnea Research LLC
// Copyright (c) 2009-2011 Simen Svale Skogsrud
// Copyright (c) 2011 Jens Geisler
// Copyright (c) 2018 -	Bart Dring
// Use of this source code is governed by a GPLv3 license that can be found in the LICENSE file.

/*
  Planner.cpp - buffers movement commands and manages the acceleration profile plan
*/

#include "Planner.h"
#include "Machine/MachineConfig.h"

#include <cstdlib>  // PSoc Required for labs
#include <cmath>

static plan_block_t* block_buffer = nullptr;  // A ring buffer for motion instructions
static uint8_t       block_buffer_tail;       // Index of the block to process now
static uint8_t       block_buffer_head;       // Index of the next block to be pushed
static uint8_t       next_buffer_head;        // Index of the next buffer head

This file has been truncated. show original

Stoli · March 17, 2025, 7:09pm

Thanks. I’m in the middle of trying out some AI generated code that I hope will get me in the ballpark.

In the meantime, I’m trying to figure out the fluidNC units for accel. The default in my config.yaml is 200 mm/sec^2. The Copilot generated code specifies it wants mm/min^2 and its example uses 300 mm/min^2. I think 200mm/sec^2 converts to 720,000 mm/min^2. Seems kind of big. Maybe it’s a typo by Copilot?

jeyeager · March 17, 2025, 7:15pm

That checks out.

Stoli · March 17, 2025, 7:16pm

I thought I had a handle on this until I read the comments in your reference. Now I’m really confused.

jeyeager · March 17, 2025, 7:28pm

Yea, the planner is literally the most complicated part of GRBL/FluidNC. I don’t pretend to understand it other than at a 10,000 foot level.

When I saw this thread, I had 2 thoughts:

Good. I have no idea how long my plots are going to take.
It’s probably easier to just crank up the acceleration. If the feedrate stays low, I’m not sure how much that affects the experience of the plot. At high feedrates, acceleration has a pretty significant impact.

Stoli · March 17, 2025, 8:03pm

Yeah, I don’t think we need to get in the weeds here. I think if we look at the difference between the entry and exit vector at each point in the gcode and apply a simple decel/accel formula we should be close. For instance, if the difference is 0, then there is no change in velocity as it traverses the point. If it is 90 degrees or more, then the velocity slows to 0 based on the acceleration set in the config.yaml and accelerates back to the F speed accordingly. And, maybe apply a cosine relationship to the velocity for 0 to 90 degree changes.

i.e. velocity at the point = Fspeed * cos(angular change)