Adjust indentation in ubl.h
This commit is contained in:
parent
f5c5c1f771
commit
4abd53f65a
514
Marlin/ubl.h
514
Marlin/ubl.h
|
@ -20,320 +20,322 @@
|
|||
*
|
||||
*/
|
||||
|
||||
#include "Marlin.h"
|
||||
#include "math.h"
|
||||
#include "vector_3.h"
|
||||
|
||||
#ifndef UNIFIED_BED_LEVELING_H
|
||||
#define UNIFIED_BED_LEVELING_H
|
||||
|
||||
#if ENABLED(AUTO_BED_LEVELING_UBL)
|
||||
#include "MarlinConfig.h"
|
||||
|
||||
#define UBL_VERSION "1.00"
|
||||
#define UBL_OK false
|
||||
#define UBL_ERR true
|
||||
#if ENABLED(AUTO_BED_LEVELING_UBL)
|
||||
|
||||
typedef struct {
|
||||
int8_t x_index, y_index;
|
||||
float distance; // When populated, the distance from the search location
|
||||
} mesh_index_pair;
|
||||
#include "Marlin.h"
|
||||
#include "math.h"
|
||||
#include "vector_3.h"
|
||||
|
||||
enum MeshPointType { INVALID, REAL, SET_IN_BITMAP };
|
||||
#define UBL_VERSION "1.00"
|
||||
#define UBL_OK false
|
||||
#define UBL_ERR true
|
||||
|
||||
void dump(char * const str, const float &f);
|
||||
bool ubl_lcd_clicked();
|
||||
void probe_entire_mesh(const float&, const float&, const bool, const bool, const bool);
|
||||
void debug_current_and_destination(char *title);
|
||||
void ubl_line_to_destination(const float&, uint8_t);
|
||||
void manually_probe_remaining_mesh(const float&, const float&, const float&, const float&, const bool);
|
||||
vector_3 tilt_mesh_based_on_3pts(const float&, const float&, const float&);
|
||||
float measure_business_card_thickness(const float&);
|
||||
mesh_index_pair find_closest_mesh_point_of_type(const MeshPointType, const float&, const float&, const bool, unsigned int[16], bool);
|
||||
void find_mean_mesh_height();
|
||||
void shift_mesh_height();
|
||||
bool g29_parameter_parsing();
|
||||
void g29_what_command();
|
||||
void g29_eeprom_dump();
|
||||
void g29_compare_current_mesh_to_stored_mesh();
|
||||
void fine_tune_mesh(const float&, const float&, const bool);
|
||||
void bit_clear(uint16_t bits[16], uint8_t x, uint8_t y);
|
||||
void bit_set(uint16_t bits[16], uint8_t x, uint8_t y);
|
||||
bool is_bit_set(uint16_t bits[16], uint8_t x, uint8_t y);
|
||||
char *ftostr43sign(const float&, char);
|
||||
typedef struct {
|
||||
int8_t x_index, y_index;
|
||||
float distance; // When populated, the distance from the search location
|
||||
} mesh_index_pair;
|
||||
|
||||
void gcode_G26();
|
||||
void gcode_G28();
|
||||
void gcode_G29();
|
||||
extern char conv[9];
|
||||
enum MeshPointType { INVALID, REAL, SET_IN_BITMAP };
|
||||
|
||||
void save_ubl_active_state_and_disable();
|
||||
void restore_ubl_active_state_and_leave();
|
||||
void dump(char * const str, const float &f);
|
||||
bool ubl_lcd_clicked();
|
||||
void probe_entire_mesh(const float&, const float&, const bool, const bool, const bool);
|
||||
void debug_current_and_destination(char *title);
|
||||
void ubl_line_to_destination(const float&, uint8_t);
|
||||
void manually_probe_remaining_mesh(const float&, const float&, const float&, const float&, const bool);
|
||||
vector_3 tilt_mesh_based_on_3pts(const float&, const float&, const float&);
|
||||
float measure_business_card_thickness(const float&);
|
||||
mesh_index_pair find_closest_mesh_point_of_type(const MeshPointType, const float&, const float&, const bool, unsigned int[16], bool);
|
||||
void find_mean_mesh_height();
|
||||
void shift_mesh_height();
|
||||
bool g29_parameter_parsing();
|
||||
void g29_what_command();
|
||||
void g29_eeprom_dump();
|
||||
void g29_compare_current_mesh_to_stored_mesh();
|
||||
void fine_tune_mesh(const float&, const float&, const bool);
|
||||
void bit_clear(uint16_t bits[16], uint8_t x, uint8_t y);
|
||||
void bit_set(uint16_t bits[16], uint8_t x, uint8_t y);
|
||||
bool is_bit_set(uint16_t bits[16], uint8_t x, uint8_t y);
|
||||
char *ftostr43sign(const float&, char);
|
||||
|
||||
///////////////////////////////////////////////////////////////////////////////////////////////////////
|
||||
void gcode_G26();
|
||||
void gcode_G28();
|
||||
void gcode_G29();
|
||||
extern char conv[9];
|
||||
|
||||
#if ENABLED(ULTRA_LCD)
|
||||
extern char lcd_status_message[];
|
||||
void lcd_quick_feedback();
|
||||
void save_ubl_active_state_and_disable();
|
||||
void restore_ubl_active_state_and_leave();
|
||||
|
||||
///////////////////////////////////////////////////////////////////////////////////////////////////////
|
||||
|
||||
#if ENABLED(ULTRA_LCD)
|
||||
extern char lcd_status_message[];
|
||||
void lcd_quick_feedback();
|
||||
#endif
|
||||
|
||||
enum MBLStatus { MBL_STATUS_NONE = 0, MBL_STATUS_HAS_MESH_BIT = 0, MBL_STATUS_ACTIVE_BIT = 1 };
|
||||
|
||||
#define MESH_X_DIST (float(UBL_MESH_MAX_X - (UBL_MESH_MIN_X)) / float(GRID_MAX_POINTS_X - 1))
|
||||
#define MESH_Y_DIST (float(UBL_MESH_MAX_Y - (UBL_MESH_MIN_Y)) / float(GRID_MAX_POINTS_Y - 1))
|
||||
|
||||
typedef struct {
|
||||
bool active = false;
|
||||
float z_offset = 0.0;
|
||||
int8_t eeprom_storage_slot = -1,
|
||||
n_x = GRID_MAX_POINTS_X,
|
||||
n_y = GRID_MAX_POINTS_Y;
|
||||
|
||||
float mesh_x_min = UBL_MESH_MIN_X,
|
||||
mesh_y_min = UBL_MESH_MIN_Y,
|
||||
mesh_x_max = UBL_MESH_MAX_X,
|
||||
mesh_y_max = UBL_MESH_MAX_Y,
|
||||
mesh_x_dist = MESH_X_DIST,
|
||||
mesh_y_dist = MESH_Y_DIST;
|
||||
|
||||
#if ENABLED(ENABLE_LEVELING_FADE_HEIGHT)
|
||||
float g29_correction_fade_height = 10.0,
|
||||
g29_fade_height_multiplier = 1.0 / 10.0; // It's cheaper to do a floating point multiply than divide,
|
||||
// so keep this value and its reciprocal.
|
||||
#endif
|
||||
|
||||
enum MBLStatus { MBL_STATUS_NONE = 0, MBL_STATUS_HAS_MESH_BIT = 0, MBL_STATUS_ACTIVE_BIT = 1 };
|
||||
// If you change this struct, adjust TOTAL_STRUCT_SIZE
|
||||
|
||||
#define MESH_X_DIST (float(UBL_MESH_MAX_X - (UBL_MESH_MIN_X)) / float(GRID_MAX_POINTS_X - 1))
|
||||
#define MESH_Y_DIST (float(UBL_MESH_MAX_Y - (UBL_MESH_MIN_Y)) / float(GRID_MAX_POINTS_Y - 1))
|
||||
#define TOTAL_STRUCT_SIZE 40 // Total size of the above fields
|
||||
|
||||
typedef struct {
|
||||
bool active = false;
|
||||
float z_offset = 0.0;
|
||||
int8_t eeprom_storage_slot = -1,
|
||||
n_x = GRID_MAX_POINTS_X,
|
||||
n_y = GRID_MAX_POINTS_Y;
|
||||
// padding provides space to add state variables without
|
||||
// changing the location of data structures in the EEPROM.
|
||||
// This is for compatibility with future versions to keep
|
||||
// users from having to regenerate their mesh data.
|
||||
unsigned char padding[64 - TOTAL_STRUCT_SIZE];
|
||||
|
||||
float mesh_x_min = UBL_MESH_MIN_X,
|
||||
mesh_y_min = UBL_MESH_MIN_Y,
|
||||
mesh_x_max = UBL_MESH_MAX_X,
|
||||
mesh_y_max = UBL_MESH_MAX_Y,
|
||||
mesh_x_dist = MESH_X_DIST,
|
||||
mesh_y_dist = MESH_Y_DIST;
|
||||
} ubl_state;
|
||||
|
||||
#if ENABLED(ENABLE_LEVELING_FADE_HEIGHT)
|
||||
float g29_correction_fade_height = 10.0,
|
||||
g29_fade_height_multiplier = 1.0 / 10.0; // It's cheaper to do a floating point multiply than divide,
|
||||
// so keep this value and its reciprocal.
|
||||
#endif
|
||||
class unified_bed_leveling {
|
||||
private:
|
||||
|
||||
// If you change this struct, adjust TOTAL_STRUCT_SIZE
|
||||
static float last_specified_z;
|
||||
|
||||
#define TOTAL_STRUCT_SIZE 40 // Total size of the above fields
|
||||
public:
|
||||
|
||||
// padding provides space to add state variables without
|
||||
// changing the location of data structures in the EEPROM.
|
||||
// This is for compatibility with future versions to keep
|
||||
// users from having to regenerate their mesh data.
|
||||
unsigned char padding[64 - TOTAL_STRUCT_SIZE];
|
||||
static ubl_state state, pre_initialized;
|
||||
|
||||
} ubl_state;
|
||||
static float z_values[GRID_MAX_POINTS_X][GRID_MAX_POINTS_Y],
|
||||
mesh_index_to_xpos[GRID_MAX_POINTS_X + 1], // +1 safety margin for now, until determinism prevails
|
||||
mesh_index_to_ypos[GRID_MAX_POINTS_Y + 1];
|
||||
|
||||
class unified_bed_leveling {
|
||||
private:
|
||||
static bool g26_debug_flag,
|
||||
has_control_of_lcd_panel;
|
||||
|
||||
static float last_specified_z;
|
||||
static int8_t eeprom_start;
|
||||
|
||||
public:
|
||||
static volatile int encoder_diff; // Volatile because it's changed at interrupt time.
|
||||
|
||||
static ubl_state state, pre_initialized;
|
||||
unified_bed_leveling();
|
||||
|
||||
static float z_values[GRID_MAX_POINTS_X][GRID_MAX_POINTS_Y],
|
||||
mesh_index_to_xpos[GRID_MAX_POINTS_X + 1], // +1 safety margin for now, until determinism prevails
|
||||
mesh_index_to_ypos[GRID_MAX_POINTS_Y + 1];
|
||||
static void display_map(const int);
|
||||
|
||||
static bool g26_debug_flag,
|
||||
has_control_of_lcd_panel;
|
||||
static void reset();
|
||||
static void invalidate();
|
||||
|
||||
static int8_t eeprom_start;
|
||||
static void store_state();
|
||||
static void load_state();
|
||||
static void store_mesh(const int16_t);
|
||||
static void load_mesh(const int16_t);
|
||||
|
||||
static volatile int encoder_diff; // Volatile because it's changed at interrupt time.
|
||||
static bool sanity_check();
|
||||
|
||||
unified_bed_leveling();
|
||||
static FORCE_INLINE void set_z(const int8_t px, const int8_t py, const float &z) { z_values[px][py] = z; }
|
||||
|
||||
static void display_map(const int);
|
||||
static int8_t get_cell_index_x(const float &x) {
|
||||
const int8_t cx = (x - (UBL_MESH_MIN_X)) * (1.0 / (MESH_X_DIST));
|
||||
return constrain(cx, 0, (GRID_MAX_POINTS_X) - 1); // -1 is appropriate if we want all movement to the X_MAX
|
||||
} // position. But with this defined this way, it is possible
|
||||
// to extrapolate off of this point even further out. Probably
|
||||
// that is OK because something else should be keeping that from
|
||||
// happening and should not be worried about at this level.
|
||||
static int8_t get_cell_index_y(const float &y) {
|
||||
const int8_t cy = (y - (UBL_MESH_MIN_Y)) * (1.0 / (MESH_Y_DIST));
|
||||
return constrain(cy, 0, (GRID_MAX_POINTS_Y) - 1); // -1 is appropriate if we want all movement to the Y_MAX
|
||||
} // position. But with this defined this way, it is possible
|
||||
// to extrapolate off of this point even further out. Probably
|
||||
// that is OK because something else should be keeping that from
|
||||
// happening and should not be worried about at this level.
|
||||
|
||||
static void reset();
|
||||
static void invalidate();
|
||||
static int8_t find_closest_x_index(const float &x) {
|
||||
const int8_t px = (x - (UBL_MESH_MIN_X) + (MESH_X_DIST) * 0.5) * (1.0 / (MESH_X_DIST));
|
||||
return WITHIN(px, 0, GRID_MAX_POINTS_X - 1) ? px : -1;
|
||||
}
|
||||
|
||||
static void store_state();
|
||||
static void load_state();
|
||||
static void store_mesh(const int16_t);
|
||||
static void load_mesh(const int16_t);
|
||||
static int8_t find_closest_y_index(const float &y) {
|
||||
const int8_t py = (y - (UBL_MESH_MIN_Y) + (MESH_Y_DIST) * 0.5) * (1.0 / (MESH_Y_DIST));
|
||||
return WITHIN(py, 0, GRID_MAX_POINTS_Y - 1) ? py : -1;
|
||||
}
|
||||
|
||||
static bool sanity_check();
|
||||
/**
|
||||
* z2 --|
|
||||
* z0 | |
|
||||
* | | + (z2-z1)
|
||||
* z1 | | |
|
||||
* ---+-------------+--------+-- --|
|
||||
* a1 a0 a2
|
||||
* |<---delta_a---------->|
|
||||
*
|
||||
* calc_z0 is the basis for all the Mesh Based correction. It is used to
|
||||
* find the expected Z Height at a position between two known Z-Height locations.
|
||||
*
|
||||
* It is fairly expensive with its 4 floating point additions and 2 floating point
|
||||
* multiplications.
|
||||
*/
|
||||
static FORCE_INLINE float calc_z0(const float &a0, const float &a1, const float &z1, const float &a2, const float &z2) {
|
||||
return z1 + (z2 - z1) * (a0 - a1) / (a2 - a1);
|
||||
}
|
||||
|
||||
static FORCE_INLINE void set_z(const int8_t px, const int8_t py, const float &z) { z_values[px][py] = z; }
|
||||
|
||||
static int8_t get_cell_index_x(const float &x) {
|
||||
const int8_t cx = (x - (UBL_MESH_MIN_X)) * (1.0 / (MESH_X_DIST));
|
||||
return constrain(cx, 0, (GRID_MAX_POINTS_X) - 1); // -1 is appropriate if we want all movement to the X_MAX
|
||||
} // position. But with this defined this way, it is possible
|
||||
// to extrapolate off of this point even further out. Probably
|
||||
// that is OK because something else should be keeping that from
|
||||
// happening and should not be worried about at this level.
|
||||
static int8_t get_cell_index_y(const float &y) {
|
||||
const int8_t cy = (y - (UBL_MESH_MIN_Y)) * (1.0 / (MESH_Y_DIST));
|
||||
return constrain(cy, 0, (GRID_MAX_POINTS_Y) - 1); // -1 is appropriate if we want all movement to the Y_MAX
|
||||
} // position. But with this defined this way, it is possible
|
||||
// to extrapolate off of this point even further out. Probably
|
||||
// that is OK because something else should be keeping that from
|
||||
// happening and should not be worried about at this level.
|
||||
|
||||
static int8_t find_closest_x_index(const float &x) {
|
||||
const int8_t px = (x - (UBL_MESH_MIN_X) + (MESH_X_DIST) * 0.5) * (1.0 / (MESH_X_DIST));
|
||||
return WITHIN(px, 0, GRID_MAX_POINTS_X - 1) ? px : -1;
|
||||
/**
|
||||
* z_correction_for_x_on_horizontal_mesh_line is an optimization for
|
||||
* the rare occasion when a point lies exactly on a Mesh line (denoted by index yi).
|
||||
*/
|
||||
static inline float z_correction_for_x_on_horizontal_mesh_line(const float &lx0, const int x1_i, const int yi) {
|
||||
if (!WITHIN(x1_i, 0, GRID_MAX_POINTS_X - 1) || !WITHIN(yi, 0, GRID_MAX_POINTS_Y - 1)) {
|
||||
SERIAL_ECHOPAIR("? in z_correction_for_x_on_horizontal_mesh_line(lx0=", lx0);
|
||||
SERIAL_ECHOPAIR(",x1_i=", x1_i);
|
||||
SERIAL_ECHOPAIR(",yi=", yi);
|
||||
SERIAL_CHAR(')');
|
||||
SERIAL_EOL;
|
||||
return NAN;
|
||||
}
|
||||
|
||||
static int8_t find_closest_y_index(const float &y) {
|
||||
const int8_t py = (y - (UBL_MESH_MIN_Y) + (MESH_Y_DIST) * 0.5) * (1.0 / (MESH_Y_DIST));
|
||||
return WITHIN(py, 0, GRID_MAX_POINTS_Y - 1) ? py : -1;
|
||||
const float xratio = (RAW_X_POSITION(lx0) - mesh_index_to_xpos[x1_i]) * (1.0 / (MESH_X_DIST)),
|
||||
z1 = z_values[x1_i][yi];
|
||||
|
||||
return z1 + xratio * (z_values[x1_i + 1][yi] - z1);
|
||||
}
|
||||
|
||||
//
|
||||
// See comments above for z_correction_for_x_on_horizontal_mesh_line
|
||||
//
|
||||
static inline float z_correction_for_y_on_vertical_mesh_line(const float &ly0, const int xi, const int y1_i) {
|
||||
if (!WITHIN(xi, 0, GRID_MAX_POINTS_X - 1) || !WITHIN(y1_i, 0, GRID_MAX_POINTS_Y - 1)) {
|
||||
SERIAL_ECHOPAIR("? in get_z_correction_along_vertical_mesh_line_at_specific_x(ly0=", ly0);
|
||||
SERIAL_ECHOPAIR(", x1_i=", xi);
|
||||
SERIAL_ECHOPAIR(", yi=", y1_i);
|
||||
SERIAL_CHAR(')');
|
||||
SERIAL_EOL;
|
||||
return NAN;
|
||||
}
|
||||
|
||||
/**
|
||||
* z2 --|
|
||||
* z0 | |
|
||||
* | | + (z2-z1)
|
||||
* z1 | | |
|
||||
* ---+-------------+--------+-- --|
|
||||
* a1 a0 a2
|
||||
* |<---delta_a---------->|
|
||||
*
|
||||
* calc_z0 is the basis for all the Mesh Based correction. It is used to
|
||||
* find the expected Z Height at a position between two known Z-Height locations.
|
||||
*
|
||||
* It is fairly expensive with its 4 floating point additions and 2 floating point
|
||||
* multiplications.
|
||||
*/
|
||||
static FORCE_INLINE float calc_z0(const float &a0, const float &a1, const float &z1, const float &a2, const float &z2) {
|
||||
return z1 + (z2 - z1) * (a0 - a1) / (a2 - a1);
|
||||
}
|
||||
const float yratio = (RAW_Y_POSITION(ly0) - mesh_index_to_ypos[y1_i]) * (1.0 / (MESH_Y_DIST)),
|
||||
z1 = z_values[xi][y1_i];
|
||||
|
||||
/**
|
||||
* z_correction_for_x_on_horizontal_mesh_line is an optimization for
|
||||
* the rare occasion when a point lies exactly on a Mesh line (denoted by index yi).
|
||||
*/
|
||||
static inline float z_correction_for_x_on_horizontal_mesh_line(const float &lx0, const int x1_i, const int yi) {
|
||||
if (!WITHIN(x1_i, 0, GRID_MAX_POINTS_X - 1) || !WITHIN(yi, 0, GRID_MAX_POINTS_Y - 1)) {
|
||||
SERIAL_ECHOPAIR("? in z_correction_for_x_on_horizontal_mesh_line(lx0=", lx0);
|
||||
SERIAL_ECHOPAIR(",x1_i=", x1_i);
|
||||
SERIAL_ECHOPAIR(",yi=", yi);
|
||||
SERIAL_CHAR(')');
|
||||
SERIAL_EOL;
|
||||
return NAN;
|
||||
}
|
||||
return z1 + yratio * (z_values[xi][y1_i + 1] - z1);
|
||||
}
|
||||
|
||||
const float xratio = (RAW_X_POSITION(lx0) - mesh_index_to_xpos[x1_i]) * (1.0 / (MESH_X_DIST)),
|
||||
z1 = z_values[x1_i][yi];
|
||||
/**
|
||||
* This is the generic Z-Correction. It works anywhere within a Mesh Cell. It first
|
||||
* does a linear interpolation along both of the bounding X-Mesh-Lines to find the
|
||||
* Z-Height at both ends. Then it does a linear interpolation of these heights based
|
||||
* on the Y position within the cell.
|
||||
*/
|
||||
static float get_z_correction(const float &lx0, const float &ly0) {
|
||||
const int8_t cx = get_cell_index_x(RAW_X_POSITION(lx0)),
|
||||
cy = get_cell_index_y(RAW_Y_POSITION(ly0));
|
||||
|
||||
return z1 + xratio * (z_values[x1_i + 1][yi] - z1);
|
||||
}
|
||||
if (!WITHIN(cx, 0, GRID_MAX_POINTS_X - 1) || !WITHIN(cy, 0, GRID_MAX_POINTS_Y - 1)) {
|
||||
|
||||
//
|
||||
// See comments above for z_correction_for_x_on_horizontal_mesh_line
|
||||
//
|
||||
static inline float z_correction_for_y_on_vertical_mesh_line(const float &ly0, const int xi, const int y1_i) {
|
||||
if (!WITHIN(xi, 0, GRID_MAX_POINTS_X - 1) || !WITHIN(y1_i, 0, GRID_MAX_POINTS_Y - 1)) {
|
||||
SERIAL_ECHOPAIR("? in get_z_correction_along_vertical_mesh_line_at_specific_x(ly0=", ly0);
|
||||
SERIAL_ECHOPAIR(", x1_i=", xi);
|
||||
SERIAL_ECHOPAIR(", yi=", y1_i);
|
||||
SERIAL_CHAR(')');
|
||||
SERIAL_EOL;
|
||||
return NAN;
|
||||
}
|
||||
SERIAL_ECHOPAIR("? in get_z_correction(lx0=", lx0);
|
||||
SERIAL_ECHOPAIR(", ly0=", ly0);
|
||||
SERIAL_CHAR(')');
|
||||
SERIAL_EOL;
|
||||
|
||||
const float yratio = (RAW_Y_POSITION(ly0) - mesh_index_to_ypos[y1_i]) * (1.0 / (MESH_Y_DIST)),
|
||||
z1 = z_values[xi][y1_i];
|
||||
|
||||
return z1 + yratio * (z_values[xi][y1_i + 1] - z1);
|
||||
}
|
||||
|
||||
/**
|
||||
* This is the generic Z-Correction. It works anywhere within a Mesh Cell. It first
|
||||
* does a linear interpolation along both of the bounding X-Mesh-Lines to find the
|
||||
* Z-Height at both ends. Then it does a linear interpolation of these heights based
|
||||
* on the Y position within the cell.
|
||||
*/
|
||||
static float get_z_correction(const float &lx0, const float &ly0) {
|
||||
const int8_t cx = get_cell_index_x(RAW_X_POSITION(lx0)),
|
||||
cy = get_cell_index_y(RAW_Y_POSITION(ly0));
|
||||
|
||||
if (!WITHIN(cx, 0, GRID_MAX_POINTS_X - 1) || !WITHIN(cy, 0, GRID_MAX_POINTS_Y - 1)) {
|
||||
|
||||
SERIAL_ECHOPAIR("? in get_z_correction(lx0=", lx0);
|
||||
SERIAL_ECHOPAIR(", ly0=", ly0);
|
||||
SERIAL_CHAR(')');
|
||||
SERIAL_EOL;
|
||||
|
||||
#if ENABLED(ULTRA_LCD)
|
||||
strcpy(lcd_status_message, "get_z_correction() indexes out of range.");
|
||||
lcd_quick_feedback();
|
||||
#endif
|
||||
return 0.0; // this used to return state.z_offset
|
||||
}
|
||||
|
||||
const float z1 = calc_z0(RAW_X_POSITION(lx0),
|
||||
mesh_index_to_xpos[cx], z_values[cx][cy],
|
||||
mesh_index_to_xpos[cx + 1], z_values[cx + 1][cy]),
|
||||
z2 = calc_z0(RAW_X_POSITION(lx0),
|
||||
mesh_index_to_xpos[cx], z_values[cx][cy + 1],
|
||||
mesh_index_to_xpos[cx + 1], z_values[cx + 1][cy + 1]);
|
||||
float z0 = calc_z0(RAW_Y_POSITION(ly0),
|
||||
mesh_index_to_ypos[cy], z1,
|
||||
mesh_index_to_ypos[cy + 1], z2);
|
||||
|
||||
#if ENABLED(DEBUG_LEVELING_FEATURE)
|
||||
if (DEBUGGING(MESH_ADJUST)) {
|
||||
SERIAL_ECHOPAIR(" raw get_z_correction(", lx0);
|
||||
SERIAL_CHAR(',')
|
||||
SERIAL_ECHO(ly0);
|
||||
SERIAL_ECHOPGM(") = ");
|
||||
SERIAL_ECHO_F(z0, 6);
|
||||
}
|
||||
#if ENABLED(ULTRA_LCD)
|
||||
strcpy(lcd_status_message, "get_z_correction() indexes out of range.");
|
||||
lcd_quick_feedback();
|
||||
#endif
|
||||
return 0.0; // this used to return state.z_offset
|
||||
}
|
||||
|
||||
const float z1 = calc_z0(RAW_X_POSITION(lx0),
|
||||
mesh_index_to_xpos[cx], z_values[cx][cy],
|
||||
mesh_index_to_xpos[cx + 1], z_values[cx + 1][cy]),
|
||||
z2 = calc_z0(RAW_X_POSITION(lx0),
|
||||
mesh_index_to_xpos[cx], z_values[cx][cy + 1],
|
||||
mesh_index_to_xpos[cx + 1], z_values[cx + 1][cy + 1]);
|
||||
float z0 = calc_z0(RAW_Y_POSITION(ly0),
|
||||
mesh_index_to_ypos[cy], z1,
|
||||
mesh_index_to_ypos[cy + 1], z2);
|
||||
|
||||
#if ENABLED(DEBUG_LEVELING_FEATURE)
|
||||
if (DEBUGGING(MESH_ADJUST)) {
|
||||
SERIAL_ECHOPAIR(" raw get_z_correction(", lx0);
|
||||
SERIAL_CHAR(',')
|
||||
SERIAL_ECHO(ly0);
|
||||
SERIAL_ECHOPGM(") = ");
|
||||
SERIAL_ECHO_F(z0, 6);
|
||||
}
|
||||
#endif
|
||||
|
||||
#if ENABLED(DEBUG_LEVELING_FEATURE)
|
||||
if (DEBUGGING(MESH_ADJUST)) {
|
||||
SERIAL_ECHOPGM(" >>>---> ");
|
||||
SERIAL_ECHO_F(z0, 6);
|
||||
SERIAL_EOL;
|
||||
}
|
||||
#endif
|
||||
|
||||
if (isnan(z0)) { // if part of the Mesh is undefined, it will show up as NAN
|
||||
z0 = 0.0; // in ubl.z_values[][] and propagate through the
|
||||
// calculations. If our correction is NAN, we throw it out
|
||||
// because part of the Mesh is undefined and we don't have the
|
||||
// information we need to complete the height correction.
|
||||
|
||||
#if ENABLED(DEBUG_LEVELING_FEATURE)
|
||||
if (DEBUGGING(MESH_ADJUST)) {
|
||||
SERIAL_ECHOPGM(" >>>---> ");
|
||||
SERIAL_ECHO_F(z0, 6);
|
||||
SERIAL_ECHOPAIR("??? Yikes! NAN in get_z_correction(", lx0);
|
||||
SERIAL_CHAR(',');
|
||||
SERIAL_ECHO(ly0);
|
||||
SERIAL_CHAR(')');
|
||||
SERIAL_EOL;
|
||||
}
|
||||
#endif
|
||||
}
|
||||
return z0; // there used to be a +state.z_offset on this line
|
||||
}
|
||||
|
||||
if (isnan(z0)) { // if part of the Mesh is undefined, it will show up as NAN
|
||||
z0 = 0.0; // in ubl.z_values[][] and propagate through the
|
||||
// calculations. If our correction is NAN, we throw it out
|
||||
// because part of the Mesh is undefined and we don't have the
|
||||
// information we need to complete the height correction.
|
||||
/**
|
||||
* This function sets the Z leveling fade factor based on the given Z height,
|
||||
* only re-calculating when necessary.
|
||||
*
|
||||
* Returns 1.0 if g29_correction_fade_height is 0.0.
|
||||
* Returns 0.0 if Z is past the specified 'Fade Height'.
|
||||
*/
|
||||
#if ENABLED(ENABLE_LEVELING_FADE_HEIGHT)
|
||||
|
||||
#if ENABLED(DEBUG_LEVELING_FEATURE)
|
||||
if (DEBUGGING(MESH_ADJUST)) {
|
||||
SERIAL_ECHOPAIR("??? Yikes! NAN in get_z_correction(", lx0);
|
||||
SERIAL_CHAR(',');
|
||||
SERIAL_ECHO(ly0);
|
||||
SERIAL_CHAR(')');
|
||||
SERIAL_EOL;
|
||||
}
|
||||
#endif
|
||||
static FORCE_INLINE float fade_scaling_factor_for_z(const float &lz) {
|
||||
if (state.g29_correction_fade_height == 0.0) return 1.0;
|
||||
|
||||
static float fade_scaling_factor = 1.0;
|
||||
const float rz = RAW_Z_POSITION(lz);
|
||||
if (last_specified_z != rz) {
|
||||
last_specified_z = rz;
|
||||
fade_scaling_factor =
|
||||
rz < state.g29_correction_fade_height
|
||||
? 1.0 - (rz * state.g29_fade_height_multiplier)
|
||||
: 0.0;
|
||||
}
|
||||
return z0; // there used to be a +state.z_offset on this line
|
||||
return fade_scaling_factor;
|
||||
}
|
||||
|
||||
/**
|
||||
* This function sets the Z leveling fade factor based on the given Z height,
|
||||
* only re-calculating when necessary.
|
||||
*
|
||||
* Returns 1.0 if g29_correction_fade_height is 0.0.
|
||||
* Returns 0.0 if Z is past the specified 'Fade Height'.
|
||||
*/
|
||||
#if ENABLED(ENABLE_LEVELING_FADE_HEIGHT)
|
||||
#endif
|
||||
|
||||
static FORCE_INLINE float fade_scaling_factor_for_z(const float &lz) {
|
||||
if (state.g29_correction_fade_height == 0.0) return 1.0;
|
||||
}; // class unified_bed_leveling
|
||||
|
||||
static float fade_scaling_factor = 1.0;
|
||||
const float rz = RAW_Z_POSITION(lz);
|
||||
if (last_specified_z != rz) {
|
||||
last_specified_z = rz;
|
||||
fade_scaling_factor =
|
||||
rz < state.g29_correction_fade_height
|
||||
? 1.0 - (rz * state.g29_fade_height_multiplier)
|
||||
: 0.0;
|
||||
}
|
||||
return fade_scaling_factor;
|
||||
}
|
||||
extern unified_bed_leveling ubl;
|
||||
|
||||
#endif
|
||||
#define UBL_LAST_EEPROM_INDEX (E2END - sizeof(unified_bed_leveling::state))
|
||||
|
||||
}; // class unified_bed_leveling
|
||||
|
||||
extern unified_bed_leveling ubl;
|
||||
|
||||
#define UBL_LAST_EEPROM_INDEX (E2END - sizeof(unified_bed_leveling::state))
|
||||
|
||||
#endif // AUTO_BED_LEVELING_UBL
|
||||
#endif // AUTO_BED_LEVELING_UBL
|
||||
#endif // UNIFIED_BED_LEVELING_H
|
||||
|
|
Loading…
Reference in a new issue