docs: Suggested improvements to Input Shaper tuning guide

Signed-off-by: Dmitry Butyugin <dmbutyugin@google.com>
This commit is contained in:
Dmitry Butyugin 2020-07-28 01:37:33 +02:00 committed by KevinOConnor
parent b944666bb1
commit f3b980c1b2
8 changed files with 14842 additions and 18216 deletions

View File

@ -162,6 +162,6 @@ consider calibrating [pressure advance](Pressure_Advance.md).
It may be necessary to perform other types of detailed printer
calibration - a number of guides are available online to help with
this (for example, do a web search for "3d printer calibration").
As an example, if you experience the effect called ghosting or ringing,
you may try following [ghosting compensation](Ghosting_Compensation.md)
As an example, if you experience the effect called ringing,
you may try following [resonance compensation](Resonance_Compensation.md)
tuning guide.

View File

@ -38,6 +38,8 @@ communication with the Klipper developers.
sensorless homing.
- [Skew correction](skew_correction.md): Adjustments for axes not
perfectly square.
- [Resonance compensation](Resonance_Compensation.md): Reduce ringing in prints
by measuring printer resonances and configuring the input shaping.
- [G-Codes](G-Codes.md): Information on commands supported by Klipper.
# Developer Documentation

View File

@ -1,9 +1,18 @@
Ghosting Compensation
Resonance Compensation
====================
Klipper supports Input Shaping - a technique that can be used to reduce ghosting
in prints. [Input shaping](https://en.wikipedia.org/wiki/Input_shaping) is an
open-loop control technique which creates a commanding signal that cancels its
Klipper supports Input Shaping - a technique that can be used to reduce ringing
(also known as echoing, ghosting or rippling) in prints. Ringing is a surface
printing defect when, typically, elements like edges repeat themselves on a
printed surface as a subtle 'echo':
|![Ringing test](img/ringing-test.jpg) |![3D Benchy](img/ringing-3dbenchy.jpg) |
|:--:|:--:|
Ringing is caused by mechanical vibrations in the printer due to quick changes
of the printing direction.
[Input shaping](https://en.wikipedia.org/wiki/Input_shaping) is an open-loop
control technique which creates a commanding signal that cancels its
own vibrations.
**Warning**: Input Shaping support is experimental. You should consider using it
@ -11,7 +20,7 @@ only if you already have ghosting and ringing in prints, otherwise it is not
advised to enable it. Input shaping requires some tuning and measurements
before it can be enabled.
Note that ghosting usually has mechanical origins: insufficiently rigid printer
Note that ringing usually has mechanical origins: insufficiently rigid printer
frame, non-tight or too springy belts, alignment issues of mechanical parts,
heavy moving mass, etc. Those should be checked and fixed first.
@ -36,7 +45,9 @@ Slice the ringing test model, which can be found in
First, measure the **ringing frequency**.
1. Increase `max_accel` and `max_accel_to_decel` parameters in your
`printer.cfg` to 7000.
`printer.cfg` to 7000. Note that this is only needed for tuning, and more
proper value will be selected in the corresponding
[section](#selecting-max_accel).
2. Restart the firmware: `RESTART`.
3. Disable Pressure Advance: `SET_PRESSURE_ADVANCE ADVANCE=0`.
4. If you have already added `[input_shaper]` section to the printer.cfg,
@ -192,16 +203,36 @@ max_accel should be chosen such as to prevent that. Another parameter that can
impact smoothing is square_corner_velocity, so it is not advisable to increase
it above the default 5 mm/sec to prevent increased smoothing.
Inspect the model for the chosen input shaper. Take a note at which
acceleration:
In order to select a suitable max_accel value, inspect the model for the chosen
input shaper. First, take a note at which acceleration ringing is still small -
that you are comfortable with it.
* ringing is still small - that you are comfortable with it;
* corners of the test model still show 'bulges' from disabled pressure
advance - remember that **input shaping is not a replacement of PA**.
Next, check the smoothing. To help with that, the test model has a small gap
in the wall (0.15 mm):
![Test gap](img/smoothing-test.png)
As the acceleration increases, so does the smoothing, and the actual gap
widens:
![Shaper smoothing](img/shaper-smoothing.jpg)
In this picture, the acceleration increases left to right, and the gap starts
to grow starting from 3500 mm/sec^2 (5-th band from the left). So the good
value for max_accel = 3000 (mm/sec^2) in this case to avoid the excessive
smoothing.
Note the acceleration when the gap is still very small in your test print.
If you see bulges, but no gap in the wall at all, even at high accelerations,
it may be due to disabled Pressure Advance, especially on Bowden extruders.
If that is the case, you may need to repeat the print with the PA enabled.
It may also be a result of a miscalibrated (too high) filament flow, so it is
a good idea to check that too.
Choose the minimum out of the two acceleration values (from ringing and
smoothing), and put it as max_accel into printer.cfg (you can delete
max_accel_or_decel or revert it to the old value).
Choose the minimum out of the two acceleration values, and put it as max_accel
into printer.cfg (you can delete max_accel_or_decel or revert it to the old
value).
As a note, it may happen - especially at low ringing frequencies - that EI
shaper will cause too much smoothing even at lower accelerations. In this case,
@ -269,6 +300,9 @@ After both new `shaper_freq_x` and `shaper_freq_y` parameters have been
calculated, you can update `[input_shaper]` section in `printer.cfg` with the
new `shaper_freq_x` and `shaper_freq_y` values.
Do not forget to revert the changes to `max_accel` and `max_accel_to_decel`
parameters in the `printer.cfg` after finishing this section.
## Pressure Advance
If you use Pressure Advance, it may need to be re-tuned. Follow the
@ -355,7 +389,7 @@ Troubleshooting and FAQ
### I cannot get reliable measurements of resonance frequencies
First, make sure it is not some other problem with the printer instead of
ghosting. If the measurements are not reliable because, say, the distance
ringing. If the measurements are not reliable because, say, the distance
between the oscillations is not stable, it might mean that the printer has
several resonance frequencies on the same axis. One may try to follow the
tuning process described in
@ -370,7 +404,7 @@ increase square_corner_velocity parameters. It might also be better to choose
MZV or even ZV input shapers over EI (or 2HUMP_EI and 3HUMP_EI shapers).
### After successfully printing for some time without ghosting, it appears to come back
### After successfully printing for some time without ringing, it appears to come back
It is possible that after some time the resonance frequencies have changed.
E.g. maybe the belts tension has changed (belts got more loose), etc. It is a
@ -393,6 +427,12 @@ SET_INPUT_SHAPER SHAPER_FREQ_X=... SHAPER_FREQ_Y=...
And similarly when switching back to carriage 0.
### Does input_shaper affect print time?
No, `input_shaper` feature has pretty much no impact on the print times by
itself. However, the value of `max_accel` certainly does (tuning of this
parameter described in [this section](#selecting-max_accel)).
Technical details
=================

Binary file not shown.

After

Width:  |  Height:  |  Size: 49 KiB

BIN
docs/img/ringing-test.jpg Normal file

Binary file not shown.

After

Width:  |  Height:  |  Size: 65 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 42 KiB

BIN
docs/img/smoothing-test.png Normal file

Binary file not shown.

After

Width:  |  Height:  |  Size: 46 KiB

File diff suppressed because it is too large Load Diff