Homing Inconsistent Duet 3

tobias_munich

Update:
I played around and changed the order of homing x and y in my homall.g
it was ok for 2 days. But today in the morning I had the problem again.
would say in 95% of the cases its ok and randomly it fails.

homall.g definitely has a problem.this macro looses lines to execute.

next step; I will setup a separat macro with the same content. lets see what's happend
surprise, surprise.

dc42

I think this needs attention from @chrishamm because as far as i can see, only users running DSF+RPi are experiencing this problem. But here is a wild guess: what happens if you put another command between each pair of M98 calls, for example M400?

T3P3Tony

another thing to look at is the DSF log when this happens, it may shed some light.

if you do

sudo journalctl -u duetcontrolserver -f

In a console on the Pi you will see the latest log entries for duet control server. As a macro is called and expanded each command will show in the logs. It will be helpful to see if there is a difference between when a macro works and it does not.

spllg

@projectR3D said in Homing Inconsistent Duet 3:

M201

i had a similar issue. decreasing x and y acceleration (M201) helped.

Garfield

I run a Pi 4, when I've encountered this and when looking for anything useful in the journals the cupboard is bare, the control server is basically disconnecting from the duet it seems, it doesn't see the problem and logs no error. My theory is that it sees no response from the Duet and remains 'busy' which locks out any other commands, it never times out but it sees no error since it hasn't actually crashed.

The only way to recover is to either reboot or use sudo systemctl restart duetcontrolserver in a terminal window, then the system will respond again, like the OP most of the times I have seen this is during homing operations which aren't particularly complex in communication terms. This works not because the control server has failed but because the item holding / excluding other actions is forcibly removed from the queue.

This behaviour is random and unpredictable and I've yet to find a trigger, we could do with a way to make the control server reporting a little more verbose - i.e to log all commands sent, responses received - to log the conversation.

T3P3Tony

@Garfield ok, yes if DCS hangs then it does not log. This is a different issue to one we had much earlier on i think. It would be worth trying what @dc42 suggested and putting a M400 between macro calls to see if that solves the issue.

How deep are your macros nested, like homeall->homez->deployprobe->(somecustommacro)->anothercustommacro...?

Garfield

I have no macro nesting at all currently my homing files are all separate with no M98 calls, I don't use deployprobe (don't see such a macro anyway or knowingly call it).

I'm not sure the DCS is hanging - p[retty sure that when I ran sudo systemctl status duetcontrolserver it comes back as 'running' but next time it does this I shall confirm exactly what status returns.

I shall also try the M400 (once I've seen the error again - problem is that I can't reproduce it at will).

tobias_munich

@dc42 (M400) thats a good point. will test it.
and good to know, that you have it on your list, many thanks.

but again. it's not a major issue. the homing scripts are working independently very well

Luke'sLaboratory

I'm a Duet 3 user, with SBC and single 3HC, 3.01 RC3, DSF 1.2.4

My homing is now consistently bugged when using the homeall.g command, regardless of whether I call individual macros or not.

It happens on my X-axis, and behaves like the X Limit Switch is pressed, immediately stopping when X is homed.

I have verified using the machine-specific diagnostics that the trigger is not being pressed, and I've confirmed that it does register correctly when I use my homex.g command and I can see it press, as well as manually pressing it.

I've tried M400 as well as changing homeall.g to incorporate the gcode from homeX, same result, it behaves like the button is always pressed but then processes homeY+homeZ appropriately.

Something weird is afoot here!

Danal

@Luke-sLaboratory said in Homing Inconsistent Duet 3:

I'm a Duet 3 user, with SBC and single 3HC, 3.01 RC3, DSF 1.2.4

My homing is now consistently bugged when using the homeall.g command, regardless of whether I call individual macros or not.

It happens on my X-axis, and behaves like the X Limit Switch is pressed, immediately stopping when X is homed.

I have verified using the machine-specific diagnostics that the trigger is not being pressed, and I've confirmed that it does register correctly when I use my homex.g command and I can see it press, as well as manually pressing it.

I've tried M400 as well as changing homeall.g to incorporate the gcode from homeX, same result, it behaves like the button is always pressed but then processes homeY+homeZ appropriately.

Something weird is afoot here!

@Luke-sLaboratory please post your config.g, and homing files.

Luke'sLaboratory

@Danal

; Drives
M569 P0.1 S1                    ; physical drive 0.1 goes forwards XY
M569 P0.0 S1                    ; physical drive 0.0 goes forwards XY
M569 P0.2 S0                    ; physical drive 0.2 goes forwards Coupler
M569 P0.3 S1				    ; physical drive 0.3 goes backwards Z
M569 P0.4 S1				    ; physical drive 0.4 goes backwards Z
M569 P0.5 S0                    ; physical drive 0.5 goes backwards Z
M569 P1.0 S0					; Extruder 1
M569 P1.1 S0					; Extruder 2
M569 P1.2 S0					; Extruder 3
M584 X0.1 Y0.0 Z0.3:0.4:0.5 C0.2 E1.0:1.1:1.2                     ; set drive mapping
M350 X16 Y16 E16:16:16 I1                      						  ; configure microstepping with interpolation
M350 Z16 C8 I0														  ; No interpolation on Z Axis
M92 X160.00 Y160.00 Z1600.00 E409.00:409.00:409.00 C100         ; set steps per mm
M566 X900 Y900 Z600 E1800:1800:1800 C2          ; set maximum instantaneous speed changes (mm/min)
M203 X50000.00 Y50000.00 Z840 E10800.00:10800:10800 C5000   ; set maximum speeds (mm/min)
M201 X2000.00 Y2000.00 Z18 E2500:2500:2500 C400      ; set accelerations (mm/s^2)
M906 X2300 Y2300 Z1800 E1100:1100:1100 C600 I60             ; set motor currents (mA) and motor idle factor in per cent
M84 S120                                      ; Set idle timeout

;Stall Detection
M915 C S3 F0 H200					; Coupler

; Axis Limits
M208 X-17 Y-15 Z0 C0 S1                             ; set axis minima
M208 X622 Y660 Z950 C500 S0                       ; set axis maxima
M671 X290.0:-32:627 Y658:185:186 S20 ; leadscrews at rear middle, front left+right

; Endstops
M574 X1 S1 P"!io2.in"                         ; configure active-low endstop for low end on X via pin io2.in
M574 Y1 S1 P"!io0.in"                         ; configure active-low endstop for low end on Y via pin io0.in
;M574 Y2 S1 P"!io1.in"			      ; configure active-low endstop for high end on Y via pin io1.in

;Tool Presence Checking
M950 J"io4.in"

; Z-Probe

M558 P5 C"io3.in" H10 F120 T3600      ; set Z probe type to simp switch and the dive height + speeds
G31 X0 Y0 Z0 P200                        ; set Z probe trigger value, offset and trigger height
;M557 X15:600 Y15:600 S30                     ; define mesh grid - Whole Bed
M557 X220:400 Y220:400 S50                     ; define mesh grid - Center Small

for my homeall

G91                     ; relative positioning
G1 H2 Z5 F600          ; lift Z relative to current position
G90
M400
M98 P"homex.g"
M400
M98 P"homey.g"
M400
M98 P"homez.g"			; this will leave it 5mm above the bed
M400
M98 P"homec.g"

HomeX.g - same thing for homey just in y - comments are to try to diagnose this issue.

M400
;M913 X40 Y40
G1 H1 X-800 F6000 ; move quickly to X axis endstop and stop there (first pass)
G1 X5 F3600       ; go back a few mm
G1 H1 X-627 F360  ; move slowly to X axis endstop once more (second pass)
M400
;M913 X100 Y100

Danal

Looks pretty much exactly like mine on the Jubilee:

; Home X Axis

G91                     ; Set relative mode
G1 X-999 F6000 H1       ; Big negative move to search for endstop
G1 X4 F600              ; Back off the endstop
G1 X-10 F600 H1         ; Find endstop again slowly
G90                     ; Set absolute mode
G1 X0 F6000

I do see an interesting difference in our config.g: Your X endstop:
M574 X1 S1 P"!io2.in" ; configure active-low endstop for low end on X via pin io2.in

Mine:
M574 X1 S1 P"io1.in" ; Set homing switch configuration X1 = low-end, S1 = active-high (NC)

I'm wondering why the invert, the "!"... Are your switches "normally closed"? If they are "normally open", I wonder if the input pin is picking up noise when the motors are moving? I'd switch the physical wiring to "normally closed"... but, as a quick test, maybe add pullup to your definition and see what happens?
M574 X1 S1 P"!^io2.in"

P.S. I notice you are using io0.in for Y endstop. Documentation says don't use 0, it is reserved for PanelDue. However, I have used it; I missed that document at first. It seemed to work just fine.

Luke'sLaboratory

@Danal

Normally Open - My RB's were wired like that and I stick with it. no other reason than that.

Not going to say its 100% not the wiring, but this is a setup that has been fully working for a few months without changes, including the io0.in, but you're right, I could move that.

This works 100% of the time I just home X, using the script. If it was mechanical/electrical, I would expect it to behave similarly with the single call.

I'll give pullup a try, but I'm not confident it will make a difference.

spllg

@Luke-sLaboratory said in Homing Inconsistent Duet 3:

M201 X2000.00 Y2000.00 Z18 E2500:2500:2500 C400

have you tried reducing accelerations?

tobias_munich

@Luke-sLaboratory my problem is exactly the same.
My first guess was that the optical endstop has a problem. I changed my Hypercube to sensoreless homing. Which works great

My homeall.g looks exactly like yours

But anyhow the issue is still the same.

Danal

@Luke-sLaboratory said in Homing Inconsistent Duet 3:

Normally Open - My RB's were wired like that and I stick with it. no other reason than that.

Yeah, and it is easy to sit here and say "Change this, change that"... but... I'd still recommend going normally closed. For one thing, it inherently detects any wiring flaws.

"Worked for months". Could be a version change of the firmware, but much more likely to be something that changed, very subtly, in the hardware. Connection becoming slightly flakey, etc.

RE: The io0... I used it for months before I found the document. So I wouldn't change it. Also, it is Y, correct? And your real problem is with X.

In summary: I certainly can't explain how it worked or what changed. Nonetheless, I'd swap to NC.

Luke'sLaboratory

@spllg

Curious - what would acceleration do for homing?
Its not actuating the switch from pure inertia, it just thinks the button is always pressed for some reason.

Luke'sLaboratory

@Danal

I hear ya, I hear ya

The problem is that N.O can't randomly close the connection, Both ends on my switches are heat shrunk, and the other ends are properly sitting in their crimped spots.

If it was NC, yeah, if i had a poor connection, it would make sense if something broke, but here wires would have to wear+arc, and that isn't the case, at least as long as I can see, nor does it show up in firmware.

I found that If i power restart the machine, the first home works correctly, but any subsequent homes do not until reset again.

We'll see what RC4 brings us.

@Luke-sLaboratory said in Homing Inconsistent Duet 3:

The problem is that N.O can't randomly close the connection, Both ends on my switches are heat shrunk, and the other ends are properly sitting in their crimped spots.

it has more to do connections breaking rather than being made; if the NO switch gets disconnected your homing will crash. If your NC switch becomes disconnected it can be detected and crashes avoided (although the common homing sequences does not seem to take advantage of that)

Danal

Bearer is correct on one of the prime underlying reasons for the industry "best practice" for NC.

Luke, in your case, I don't suspect wiring causing an issue in terms of a wire itself causing a false trigger... I've worked with you on other discords and you "do things right" on crimping and wiring and such... Instead, I suspect "noise" or a "floating wire" or "capacitance" or something similar. Can't prove it... really don't know why it used to work and works once after a power cycle... but I do know that NC puts the wiring in a "known state", namely hard pull to ground, until that switch clicks.