Firmware wishlist and priorities for Duet WiFi and Duet Ethernet
-
10. Axis hysteresis compensation. Backlash compensation
Please please please do this… without it 3d printing will never be that good without every expensive hardware..I have a number of printers of different kinematics but ALL suffer from backlash ... Seeming the only way to get rid of 99% of it is to go with ball screws on the X/Y access but that will massively increase cost, weight and inertia leading to a slow print..
Also it seems like Marlin is finally implementing this... so..
Thanks in advance
-
10. Axis hysteresis compensation. Backlash compensation
Please please please do this… without it 3d printing will never be that good without every expensive hardware..I have a number of printers of different kinematics but ALL suffer from backlash ... Seeming the only way to get rid of 99% of it is to go with ball screws on the X/Y access but that will massively increase cost, weight and inertia leading to a slow print..
Also it seems like Marlin is finally implementing this... so..
Thanks in advance
That's an unusual observation and request. Have you got any pictures or other evidence of how this backlash \ axis hysteresis manifests itself on all your printers? From a basic engineering point of view, backlash is a known phenomenon with screw driven systems and gear trains but not with belted systems, so I'm curious as to why you think that going from a (presumably) belt driven system to a screw driven system would eliminate 99% of the backlash.
-
My votes in order of priority, though may be too late:
4. Independent homing switches for multiple Z motors.
18. Ability to update PanelDue firmware via the web interface.
7. Dynamically-varying microstepping. This will allow you to configure 256x microstepping most cases, because the microstepping will be reduced automatically during high-speed moves.
10. Axis hysteresis compensation.
8. Standstill current reduction, which will allow higher motor currents to be used when the motors are moving.
-
My Vote, though probably too late:
10. Axis hysteresis compensation.
7. Dynamically-varying microstepping.
I am sure Ian knows this but … in order to make the firmware more CNC friendly i would suggest 3 additional things.
G2/G3 circular movement support
Pitch error comp on the axis (for lead screw/ball screw support)
Change M0/M1 to follow CNC convention with stop and option stop that doesn't necessarily cancel the job or turn the heaters off. Just performs a "feed hold"My 2 cents
-
G2 and G3 are already supported.
Please explain what exactly you mean by "Pitch error comp".
What exactly is the behaviour you would like to see for M0 and M1?
-
My vote for:
7. Dynamically-varying microstepping
Probe the bed multiple times at each point and take an average
-
G2 and G3 are already supported.
Please explain what exactly you mean by "Pitch error comp".
What exactly is the behaviour you would like to see for M0 and M1?
My bad on the G2/G3. I guess I never gave it a shot…
Pitch error comp is built into almost every CNC machine tool. It allows a user to deal with manufacturing defects in the lead screw and ball screws on a machine. Most machines allow a user to "adjust" where a specific point is in the axis. So a table by default might have points every 10MM. If you tell the machine to move to X10.0 but because of the lead screw it moves to 10.02, you can then provide that difference so the motion planner can get to the proper point. This is on top of what we would call steps/mm for the axis. This allows changes over the length of the axis. Not necessary for belts, can be very necessary for screws.
For the M0/M1 commands:
These are typically "pause commands" in traditional G-code. M0 will provide a feed hold command until a cycle start has been pressed. The spindle continues to turn.
M1 does the same thing but is hooked to a switch on the controller called "optional stop". When the optional stop button is on(it is latched on, not momentary) the machine will stop just like an M0 and wait for the operator to press cycle start. If the optional stop button is off the machine ignores the command and keeps running. This is very useful when proving out new CNC programs and tooling.I admit I have not used the pause and resume commands that are in the firmware now. This would just help to bring Reprap firmware to industry standard on the subtractive manufacturing side.
Hope all that makes sense!
-
Auto squaring/calibrating by homing dual x and dual y steppers individually.
I also have an mpcnc machine that at some point I am going to swap over to a duet ethernet so I can also use it for 1 meter x 1meter x 15cm prints. It has dual x, and dual y steppers. One feature that the branch of marlin has that would be great on the duet would be to automatically be able to square off the x and y axis by moving all 4 motors independently to their own endstops
-
Auto squaring/calibrating by homing dual x and dual y steppers individually.
I also have an mpcnc machine that at some point I am going to swap over to a duet ethernet so I can also use it for 1 meter x 1meter x 15cm prints. It has dual x, and dual y steppers. One feature that the branch of marlin has that would be great on the duet would be to automatically be able to square off the x and y axis by moving all 4 motors independently to their own endstops
Hmm, curious. Does it have stops at each end of the axis? Presumably for that to work, the machine would need to have a fair amount of "slop" in the linear guides so that it can twist? Or is some other mechanism employed?
-
Hmm, curious. Does it have stops at each end of the axis? Presumably for that to work, the machine would need to have a fair amount of "slop" in the linear guides so that it can twist? Or is some other mechanism employed?
I have 8 endstops, min and max for each rail. Being a meter in length, and a mixture of stainless steel and 3d printed parts, and gt2 belts, you can get up to 2cm off square if you move just one of the motors whilst keeping the other stationary. Doesnt seem to be a problem in use though, as both motors move together to keep it square. I have successfully milled aluminium on it, though usually I just mill wood or polcarbonate.
-
Hmm, curious. Does it have stops at each end of the axis? Presumably for that to work, the machine would need to have a fair amount of "slop" in the linear guides so that it can twist? Or is some other mechanism employed?
I have 8 endstops, min and max for each rail. Being a meter in length, and a mixture of stainless steel and 3d printed parts, and gt2 belts, you can get up to 2cm off square if you move just one of the motors whilst keeping the other stationary. Doesnt seem to be a problem in use though, as both motors move together to keep it square. I have successfully milled aluminium on it, though usually I just mill wood or polcarbonate.
That's fair enough - I have twin linear rails on my CoreXY which are only held by a single bolt at each end so if the belt tension isn't exactly the same both sides, it'll twist the axis. In fact, that's how I set the belt tension - push the X is to it's extreme Y position, then adjust the belt tension on one side or the other until it's square with the frame. Obviously once set, the belt tension will keep it square regardless.
Actually, on your printer, the fact that both motors move together will keep the axes parallel but not necessarily square. So for auto squaring to work, there would need to be some inherent "flex" in the system. I guess you could use M556 axis compensation but it's a manual process to set it.
-
Auto squaring/calibrating by homing dual x and dual y steppers individually.
I also have an mpcnc machine that at some point I am going to swap over to a duet ethernet so I can also use it for 1 meter x 1meter x 15cm prints. It has dual x, and dual y steppers. One feature that the branch of marlin has that would be great on the duet would be to automatically be able to square off the x and y axis by moving all 4 motors independently to their own endstops
This is possible with the Duet/RRF already. Use M584 to create two additional invisible axes (U and V assuming that you only have XYZ) like this:
M584 X0:3 Y1:4 Z3 U10 V11 E5:6:7:8:9 P3
Then when homing X, separate the 2 X motors into X and U and home them together:
M584 X0 U3
G91
G1 S1 X-300 U-300 F3000
…and then join them again:
M584 X0:3 U10
Similarly for homing Y.
-
Wow - light bulb moment. I've just realised that I can use this for homing my extruder CoreXY that sits above the hot end CoreXY. Currently I have to remember to check that they are lined up by eye before homing, as only the hot end CoreXY has end stop switches attached. This works because the extruder CoreXY has longer axes than the hot end so they don't have to be too precisely aligned but sometimes when I've been working on the printer, I forget and if the extruders are too far to the left or front, nasty things happen.
Quick question though. Presumably one has to define the other (additional) end stop switches using M574 U and V. But these switches would need to be connected to spare E stops (as there are no U and V estops labelled as such on the board). Can you point me to where I assign an En stop to an axis?
-
E0 endstop is for U and E1 endstop is for V.
-
E0 endstop is for U and E1 endstop is for V.
Brill!. I'm currently using E0 and E1 for emergency stops on axis maxima and a separate emergency stop switch but I can re-position these to use two of the estops on the Duex 5.
-
Would implementing Z axis backlash/hysteresis compensation be any easier than X/Y axes?
I am using a belt driven Z with a worm drive reduction. And as such, upon reversing travel direction, I have ~0.07mm of dead space… That's not really a problem for regular usage (one direction movement), but I would like to use mesh bed compensation eventually, and that would pose a bit of a problem... -
On a Cartesian printer, it's equally difficult on all the axes. On a delta printer, it may be harder. It partly depends on the amount of compensation needed. Less than 1 full motor step may be easy because we can probably get away with sending the extra steps all at once; but 2 or more full steps is likely to be hard because we will need to send them slowly enough for the motor to move with them, otherwise it is likely to skip steps.
What fraction of a full step (or how many full steps) does your 0.07mm correspond to?
-
I'm using CoreXY actually. With current reduction 0.07mm corresponds to 28 full steps, but it will be 7 when correct pulleys arrive…
That would really need to use jerk/accel values to not skip steps I suppose.
The way I understand it, we would need a way to determine if axis is changing direction, and then add an appropriate amount to Z move command? Or am I missing something important here? Either way, I'm not asking/hoping this will be implemented soon (or at all), just wondering, that's all.
-
The problem is that when reversing direction, the extra steps ideally need to occur and be acted on instantaneously. If you just increase the distance moved, that will correct for under-scaling caused by backlash, but not for the other print quality problems it causes.
In belt-driven X and Y axez, I think the amount of correction needed will generally be much less than a full step; which means that the extra steps can be sent rapidly at the start of the move.
-
Two suggestions, one to improve Z probing accuracy, the other to improve mesh bed leveling (background is here: https://www.duet3d.com/forum/thread.php?pid=32725).
-
A new option for G30 single-point probe commands to probe multiple points and average them to generate the final height for (X,Y). Ideally you'd add NUMBER_OF_POINTS and RADIUS to the probe command, and G30 would then probe that number of points around the given (X,Y), as well as (X,Y) itself. (Since I'm wishing, let's consider an additional option to apply a median filter to this before averaging and/or throw out the lowest and highest measurements.) I think this would increase the accuracy and repeatability of Z probes regardless of what kind of probe you're using.
-
The ability to apply a median filter to heightmap.csv to eliminate outliers (incorrect probe heights). This could prevent one or more bad points from distorting the bed map.
-
