Posts made by jrjones88

I'm not really sure what I am doing wrong.
After fixing a bad thermistor/hot end connection, I thought I was back on track getting decent print quality.
I just changed nozzles from 0.4 to 0.6 mm on a print (layer height kept the same) and the quality was drastically worse.
On the 0.4 mm nozzle prints I was getting some striations, which I thought was strange, but could live with it. With the 0.6 mm nozzle prints, it looks like layer shifting but everything physical with the printer is in order (belts are tensioned, I don't notice skipping, etc).
The blobbing is me still needing to get retraction tuned properly, but would appreciate any advice.
0.6mm Nozzle


0.4mm nozzle, same object

Printer base is a Seemecnc RostockMaxV2 (Delta)

RepRapFirmware for Duet 2 WiFi/Ethernet version 3.3 (2021-06-15 21:44:54) running on Duet WiFi 1.0 or 1.01
Board ID: 08DAM-999TL-MQ4S8-6J1FL-3SN6S-9QBRX

Config here:

; Configuration file for Duet WiFi (firmware version 3)
; executed by the firmware on start-up
;
; generated by RepRapFirmware Configuration Tool v3.3.0 on Thu Aug 05 2021 19:10:57 GMT-0700 (Pacific Daylight Time)

; General preferences
G90                                               ; send absolute coordinates...
M83                                               ; ...but relative extruder moves
M550 P"M-Automation"                              ; set printer name
M665 R138.543 L290.8 B140 H592.5 X0.714 Y1.066 Z0.000     ; Set delta radius, diagonal rod length, printable radius and homed height
M666 X-0.353 Y-0.422 Z0.775 A-0.48 B0.30          ; put your endstop adjustments here, or let auto calibration find them

; Network
M552 S1                                           ; enable network
M586 P0 S1                                        ; enable HTTP
M586 P1 S0                                        ; disable FTP
M586 P2 S0                                        ; disable Telnet

; Drives
M569 P0 S0                                        ; physical drive 0 goes backwards
M569 P1 S1                                        ; physical drive 1 goes forwards
M569 P2 S0                                        ; physical drive 2 goes backwards
M569 P3 S1                                        ; physical drive 3 goes forwards
M584 X0 Y1 Z2 E3                                  ; set drive mapping
M350 X16 Y16 Z16 E16 I1                           ; configure microstepping with interpolation
M92 X80.00 Y80.00 Z80.00 E416.78859                  ; set steps per mm
M566 X1000.00 Y1000.00 Z1000.00 E3000.00 P1         ; set maximum instantaneous speed changes (mm/min)
M203 X15000.00 Y15000.00 Z15000.00 E3000.00       ; set maximum speeds (mm/min)
M201 X8300.00 Y8300.00 Z8300.00 E10000.00       ; set accelerations (mm/s^2)
M906 X1400 Y1400 Z1400 E1400 I30                  ; set motor currents (mA) and motor idle factor in per cent
M84 S30                                           ; Set idle timeout

; Axis Limits
M208 Z-0.2 S1                                     ; set minimum Z

; Endstops
M574 X2 S1 P"xstop"                               ; configure active-high endstop for high end on X via pin xstop
M574 Y2 S1 P"ystop"                               ; configure active-high endstop for high end on Y via pin ystop
M574 Z2 S1 P"zstop"                               ; configure active-high endstop for high end on Z via pin zstop

; Z-Probe
M558 P8 C"^zprobe.in" H2.5 F300:150 A015 S0.02 T15000           ; set Z probe type to switch and the dive height + speeds
G31 P1000 X0 Y0 Z-0.04                             ; set Z probe trigger value, offset and trigger height
M557 R120 S15                                     ; define mesh grid

; Heaters
M308 S0 P"duex.e4temp" Y"thermistor" T95205 B3285 ; configure sensor 0 as thermistor on pin duex.e4temp
M950 H0 C"bedheat" T0                             ; create chamber heater output on bedheat and map it to sensor 0
M307 H0 A19.0 C18.4 D1.6 S1.00 V12.7 B0           ; disable bang-bang mode for the chamber heater and set PWM limit
M141 H0                                           ; map chamber to heater 0
M143 H0 S105                                      ; set temperature limit for heater 0 to 105C
M308 S1 P"e0temp" Y"thermistor" T108341 B4527     ; configure sensor 1 as thermistor on pin e0temp
M950 H1 C"e0heat" T1                              ; create nozzle heater output on e0heat and map it to sensor 1
M307 B0 R2.201 C212.1 D6.73 S1.00 V12.7			  ; disable bang-bang mode for heater  and set PWM limit
M143 H1 S280                                      ; set temperature limit for heater 1 to 280C
M308 S2 P"e1temp" Y"thermistor" T108341 B4527     ; configure sensor 2 as thermistor on pin e1temp
M950 H2 C"e1heat" T2                              ; create bed heater output on e1heat and map it to sensor 2
M307 H2 A190.3 C949.5 D1.2 S1.00 V12.8 B0         ; enable bang-bang mode for the bed heater and set PWM limit
M140 H2                                           ; map heated bed to heater 2
M143 H2 S125                                      ; set temperature limit for heater 2 to 125C
M308 S3 P"bedtemp" Y"thermistor"  A"Chamber Temp"  T103972 B10191 C4.825776e-7 R4700                     ; set thermistor + ADC parameters for chamber thermistor

; Fans
M950 F0 C"fan0" Q50000                            ; create fan 0 on pin fan0 and set its frequency
M106 P0 C"CPU Fan" S0.25 H0 T30                   ; set fan 0 name and value. Thermostatic control is turned on
M950 F1 C"fan1" Q0                                ; create fan 1 on pin fan1 and set its frequency
M106 P1 C"Hot End" S1 H1 T45                      ; set fan 1 name and value. Thermostatic control is turned on
M950 F2 C"fan2" Q500                              ; create fan 2 on pin fan2 and set its frequency
M106 P2 C"Chamber Heater" S0.25 H1 T35            ; set fan 2 name and value. Thermostatic control is turned on

; Tools
M563 P0 D0 H1 F0                                  ; define tool 0
G10 P0 X0 Y0 Z0                                   ; set tool 0 axis offsets
G10 P0 R0 S0                                      ; set initial tool 0 active and standby temperatures to 0C

; Custom settings
;CPU Temp compensation
M912 P0 S-2.4
;Driver Temp Flag
M305 P101 S"Drivers"

;Retraction
M207 S2.0 R0.10 F6000 T4500 Z0
M572 D0 S0

; Miscellaneous
M501                                              ; Load saved parameters from non-volatile memory
T0                                                ; Select first tool

Let me know if it is useful to have the gcode files, as they are ~12 mb in size, so too large to upload directly to the forum.

@chrishamm said in DWC2 Uploading large Gcode files results in network error:

Please try to upgrade to DWC 2.0.5. The new version no longer relies on the "axios" library for HTTP calls but uses native AJAX calls instead, so I hope this will solve your problem.

I've haven't been able to test it for a while, but I've since upgraded to 2.0.7 and while I do have a higher success rate, the odd time the upload still fails.

I've attached 3 wireshark traces of failed uploads here: https://1drv.ms/u/s!AgTPtiCYHi3hh8oNZ5d6wYz7VkdVzQ?e=pKpMUd

I seem to be unable to upload larger >3 MB with DWC2, I always get a network error.
I have no issues uploading the same file when I revert to DWC1.

I'm running 2.05 (just went from 2.04 in case that was the problem) but I have the same issues.
I know the wifi signal strength is on the 'low' side, but if I have no issues uploading in DWC1, then why can't I upload with DWC2?

Board: Duet WiFi 1.0 or 1.01
Firmware: RepRapFirmware for Duet 2 WiFi/Ethernet 2.05 (2019-12-13b1)
Duet WiFi Server Version: 1.23

M122
=== Diagnostics ===
RepRapFirmware for Duet 2 WiFi/Ethernet version 2.05 running on Duet WiFi 1.0 or 1.01
Board ID: 08DAM-999TL-MQ4S8-6J1FL-3SN6S-9QBRX
Used output buffers: 1 of 24 (10 max)
=== RTOS ===
Static ram: 25712
Dynamic ram: 92696 of which 304 recycled
Exception stack ram used: 272
Never used ram: 12088
Tasks: NETWORK(ready,628) HEAT(blocked,1232) MAIN(running,3800) IDLE(ready,160)
Owned mutexes:
=== Platform ===
Last reset 00:07:28 ago, cause: software
Last software reset at 2020-01-08 11:06, reason: User, spinning module GCodes, available RAM 12396 bytes (slot 2)
Software reset code 0x0003 HFSR 0x00000000 CFSR 0x00000000 ICSR 0x0441f000 BFAR 0xe000ed38 SP 0xffffffff Task 0x4e49414d
Error status: 0
Free file entries: 10
SD card 0 detected, interface speed: 12.0MBytes/sec
SD card longest block write time: 166.0ms, max retries 0
MCU temperature: min 32.0, current 32.6, max 34.0
Supply voltage: min 12.7, current 12.8, max 13.0, under voltage events: 0, over voltage events: 0, power good: yes
Driver 0: standstill, SG min/max not available
Driver 1: standstill, SG min/max not available
Driver 2: standstill, SG min/max not available
Driver 3: standstill, SG min/max not available
Driver 4: standstill, SG min/max not available
Date/time: 2020-01-09 10:32:44
Cache data hit count 1403011763
Slowest loop: 53.11ms; fastest: 0.06ms
I2C nak errors 0, send timeouts 0, receive timeouts 0, finishTimeouts 0, resets 0
=== Move ===
Hiccups: 0, FreeDm: 160, MinFreeDm: 160, MaxWait: 0ms
Bed compensation in use: none, comp offset 0.000
=== DDARing ===
Scheduled moves: 0, completed moves: 0, StepErrors: 0, LaErrors: 0, Underruns: 0, 0
=== Heat ===
Bed heaters = 0 -1 -1 -1, chamberHeaters = 2 -1
Heater 1 is on, I-accum = 0.0
=== GCodes ===
Segments left: 0
Stack records: 2 allocated, 0 in use
Movement lock held by null
http is idle in state(s) 0
telnet is idle in state(s) 0
file is idle in state(s) 0
serial is idle in state(s) 0
aux is idle in state(s) 0
daemon is idle in state(s) 0
queue is idle in state(s) 0
autopause is idle in state(s) 0
Code queue is empty.
=== Network ===
Slowest loop: 189.51ms; fastest: 0.01ms
Responder states: HTTP(0) HTTP(0) HTTP(0) HTTP(0) FTP(0) Telnet(0) Telnet(0)
HTTP sessions: 1 of 8
- WiFi -
Network state is running
WiFi module is connected to access point 
Failed messages: pending 0, notready 0, noresp 0
WiFi firmware version 1.23
WiFi MAC address 5c:cf:7f:2b:eb:38
WiFi Vcc 3.38, reset reason Turned on by main processor
WiFi flash size 4194304, free heap 25936
WiFi IP address 192.168.0.56
WiFi signal strength -67dBm, reconnections 0, sleep mode modem
Socket states: 0 0 0 0 0 0 0 0

I saw that in the above M122, the SD card longest block write time: 166.0ms is pretty long, but I did a few subsequent M122 and the longest block write time is 0ms

Well I think I found the main culprit, after going through and tweaking all my accels and retraction settings, at least in PrusaSlicer.
I had "Only retract when crossing perimeters" turned on, which caused a bunch of my oozing.

But getting my acceleration and travel moves better tuned also proved benificial.
I found I could bump up all the accelerations by quite a bit, as well as the travel move speed which improved stringing as well.

Still have slight gaps inbetween the gyroid layers, but that could be because it's only 5% infill.
Thanks for your input Phaedrux.

@Phaedrux said in Infill slight underextrusion:

Do you notice the same thing on other infill patterns that don't have unsupported overhangs? Like grid, or cubic?

I tried grid, and at first glance it is better, but where the extruder starts the grid pattern there is still underextrustion, but it is much harder to see. And hard to photograph:

But I think the issue is probably with my retraction settings. I did go through the 'wizard' in KISSlicer, but it still needs tweaking. Do you happen to have a link to a good guide for that? I never spotted the guide for the volumetric extrusion but it's much easier to follow than others I found.

@Phaedrux said in Infill slight underextrusion:

@jrjones88 said in Infill slight underextrusion:

I had found the max volumetric flowrate for my hot end to be ~7.5mm^3/sec

PrusaSlicer allows you to set a volumetric speed limit on the speeds tab. This would cap your movement speeds to automatically stay within the limit.

I had forgotten I had set that a while ago in Prusaslicer...

Printed it out again with the max speeds set to 45mm/sec, and I guess the volumetric speed limit was in play for the previous prints as I can't really see too much of a difference.

Thanks!
I had found the max volumetric flowrate for my hot end to be ~7.5mm^3/sec (long bowden) but hadn't calculated the max speeds for that before.

Would make my max speed 44.6 mm/s so I'll try dropping the speeds a bit.

I always thought I was printing slow because other Rostock users self reported speeds of up to 100mm/sec.

I can't seem to get the infill in my prints to come out nicely. They always are a bit underextruded.
I've calibrated the extruder, and the solid layers and perimeters print out as I expect.

Just on infill, and I've noticed it on support structures a bit too, it seems to be underextruding.
I've used PrusaSlicer, KISSlicer, Cura, and I seem to get the same results.
I thought that the issue was adaptive layer heights, but I still get it a little bit with that feature turned off.

Any tips/ideas of how to improve this?
I thought I had my speeds on the conservative side (45 mm/s perimeters, 60mm/s infill)
Some pictures:

Gcode of the print:Circle Test.gcode
Printer settings: config.g
Duet Firmware:
Board: Duet WiFi 1.0 or 1.01
Firmware: RepRapFirmware for Duet 2 WiFi/Ethernet 2.04RC1 (2019-07-14b1)
Duet WiFi Server Version: 1.23

My printer is a Rostock Max V2

@exerqtor
Where I have my printer connected to is similar, and I'm not getting any issues uploading, so I would say that -70dBm is good enough.

I have -71dBm connected through a Ubiquiti AP-AC-LR
Unifi reports a -65 dBm signal and a TX rate of 65 Mbps
I have speeds of around 200-250 kb/s as well, which I had before too located closer to an AP.

FYI my phone connected to the same AP has a self reported signal of -65dBM and a TX rate of 130 Mbps, while Unifi reports it has a -82 dBm signal and TX rate of 234 Mbps. My phone is about 4 meters further away from the AP.

@wyvern said in Yet another cast aluminum plate topic...:

Not to take over the thread, but has anyone used 3M 468MP to attach a mirror to an aluminum bed?

I hate hate hate hate hate clips or the likes, they get in the way and reduce build area.

I restrain my glass plate in 3 directions from the side. I still need to modify (improve) it since the heat cycling loosens the screws over time, but the aluminum is flat enough that after polishing it, if I lift only the glass, the aluminum sticks to it for a few seconds and is lifted up as well.
I usually have to slide the glass plate off the aluminum plate to separate them.
I was using the binder clips before but found that it warped the glass slightly because they clamped at the very edge of the glass.

Here is another place that sells off cuts of cast aluminum, out of California: http://www.sandsmachine.com/alumweb.htm

@arnix said in Suggestion for servo motor --> big delta robot:

Sorry for late reply.

To better understand my situation, am sending latest videos. One is with slower and second is on top speed.
Its without up/down movement, just positioning and extruder should position itself on X0 Y0 ( at higher speed it misses cca 10 cm ).

1.) https://ufile.io/u2cim ( just click, free slow download / open or save ).
2.) https://ufile.io/1t63o

@691175002 and @jrjones88
I think that the problem is in the gearing and lack of encoder. The material that am using at the moment aluminum but pipe wall is 2mm and pipe diameter is 2 centimeters so there should not be any problems, IMHO. As you guys point it out, 10:1 gearbox that am using could be the main problem because it has backlash of 3 degrees which is a lot, at no load. This is the stepper motor that am using :
https://www.omc-stepperonline.com/spur-gearbox/nema-23-stepper-motor-bipolar-l76mm-w-gear-raio-101-spur-gearbox-23hs30-2804s-sg10.html

High ratio belt reducer - DIY :

• i can do that, can you guys suggest some optimal ratio and motor ?
• what kind of setup would you suggest ( power supply, controller, etc. ) that can be controlled by duet ?

Yes your backlash is the issue right now. For the optimal ration and motor, that is a compromise between speed and holding power. Since you have a 1.89 Nm geared to 10:1 right now and it is too slow (you said cycle time is 1.5 seconds, desired is 1 second), your gearing needs to be a smaller ratio, say 8:1 or 5:1.

A belt reduction is a good way to reduce backlash, but depends on how much room you have. If you goto a 5:1 ratio, you might need a new motor as well, since your output torque is reduced by half on your current motor.

I don't know what your budget is (is this hobby or work/school?). If it isn't hobby I would look at Oriental Motors, they have good motors (all kinds) and drives (that accept pulsed input). But you are looking at around $500-1000 per motor/drive combination.
For example, this motor is has 0.05° backlash. https://catalog.orientalmotor.com/item/az-series-absolute-encoder-stepper-motors/az-series-85-mm-absolute-encoder-stepper-motors-dc/azm98ac-hp15

And once you reduce your backlash, I would still look at stiffening your frame (cross braces are easy to add in your case). Take a look here: https://youtu.be/p1Lrz0gPvOA
The frame is very beefy compared to the robot. This video also has some good views on the arms (very similar to the current Delta printers, but then again, delta pick and place robots have been around longer than delta 3d printers)

Could also use wood or melamine spacers.

I don't see if you have mentioned it, but another thing that is important for precision is having a rock solid base. Ideally your frame for the robot is bolted to a concrete floor, and your frame cannot have any flex.
Not having a solid base amplifies any gearing backlash by introducing extra movement, causing you to run at a reduced speed so that movement is stable.
And it's not too difficult to calculate what your max motor load would be, but you need to do some math with intertias.

@garyd9
I'm using S8 right now, I'll try again with S6.

@crynool
Not using stepper dampers, I was considering it before getting the Duet, but the Duet makes the steppers inaudible compared to the fans running.

So I changed a few things, and I found that adding a small delay before probing seemed to help quite a bit.
Under

M558 Rnnn Z probe recovery time after triggering, default zero (seconds) (RepRapFirmware 1.17 and later)

And then I was probing for a bit without a hotend and using a dial indicator instead, and that gave me better results. How I had the hot end mounted was causing a slight interference with the delta arms at the outside of the bed that I did not see. Changing how my hotend was mounted produced the same results as with the dial indicator.

I did do some test prints, and using the probe offsets didn't really help too much. The nozzle still goes low in front of the Y and Z tower, but with the probe offsets its too high in front of the X tower.
For now I'll leave the probe offsets at zero.

Picture before I changed the hot end mounting.

Picture after I changed the hot end mounting.

Although it looks worse in the picture, it's much more even, except the spots by the Y tower, and then the large part by the Z Tower. Its still better than before, since before in those areas the nozzle would drag on the build plate, and now it is at least depositing some filament in those areas.

I got similar calibration results before and after changing the hot end mounting.

Calibrated 8 factors using 13 points, deviation before 0.049 after 0.020
Diagonal 290.800, delta radius 126.528, homed height 598.400, bed radius 140.0, X -0.209°, Y -0.254°, Z 0.000°
Endstop adjustments X0.05 Y-0.08 Z0.03, tilt X0.39% Y-0.11%

This is the heightmap without probe offsets:

heightmap (after hot end change).csv

My build plate is pretty flat, its a 3/8" cast aluminum tool and jig plate, specced to a Mic6 equivalent (thickness tolerance of +/- 0.005" or 0.127 mm,) plus a glass plate on top of it.

Yes I've read Mhackneys posts before, while helpful they didn't solve my issues.
My FSR's are setup similar to his, so rocking should be minimized.

So I probed over a radius of 125mm, with 6 factor calibration, got these results:

Calibrated 6 factors using 13 points, deviation before 0.066 after 0.063
Diagonal 290.800, delta radius 140.517, homed height 671.828, bed radius 140.0, X 0.109°, Y -0.077°, Z 0.000°
Endstop adjustments X0.79 Y-1.22 Z0.43, tilt X0.00% Y0.00%

Then I applied the trigger heights as in the guide to my bed.g file with the Z probe offset being 0.004:

; 13 points, 6 factors, probing radius: 125, probe offset (0, 0)
G30 P0 X0.00 Y125.00 Z-99999 H0.077
G30 P1 X108.25 Y62.50 Z-99999 H-0.111
G30 P2 X108.25 Y-62.50 Z-99999 H0.037
G30 P3 X0.00 Y-125.00 Z-99999 H-0.111
G30 P4 X-108.25 Y-62.50 Z-99999 H0.022
G30 P5 X-108.25 Y62.50 Z-99999 H-0.131
G30 P6 X0.00 Y62.50 Z-99999 H-0.037
G30 P7 X54.13 Y31.25 Z-99999 H-0.047
G30 P8 X54.13 Y-31.25 Z-99999 H-0.059
G30 P9 X0.00 Y-62.50 Z-99999 H0.006
G30 P10 X-54.13 Y-31.25 Z-99999 H0.013
G30 P11 X-54.13 Y31.25 Z-99999 H0.04
G30 P12 X0 Y0 Z-99999 S6

And ran auto calibration again:

Calibrated 6 factors using 13 points, deviation before 0.010 after 0.007
Diagonal 290.800, delta radius 140.560, homed height 671.817, bed radius 140.0, X 0.098°, Y -0.103°, Z 0.000°
Endstop adjustments X0.80 Y-1.22 Z0.41, tilt X0.00% Y0.00%

And ran the mesh grid compensation to map the bed again, but I'm still having the same issues:

heightmap (after 6 factor w height comp).csv

@danal said in Delta - Can't get first layer flat:

Cupping or domeing is almost always a sign that diagonal rod length (M665 L) and/or delta radius (M665 R) don't match the physical reality of the printer.

Rod length is measured "joint center to joint center" preferably to 0.01 MM accuracy, and/or marked on the rods by the manufacturer. In particular, Hayden Huntley's mag-ball-end rods have correct markings. Delta Radius is VERY difficult to physically measure, and is therefore set by auto-calibration, or by making small changes and iterating until you get a "flat looking" mesh map.

There are many ways to auto-calibrate. Personally, I use 16 points and 8 factors, and perform this before EVERY print (slicer start code). But then again, I have a 600mm (24") bed. Also, I perform auto calibration several times in a row after any physical changes (even just removing and re-installing the bed). Calibrations will "converge" when repeated.

Again, if you can't get auto to work, for now, just change M665 R in small increments, re-run mesh, and see if the cup/dome increased or got better. Iterate until flat.

Either auto or manual, be sure and store calibration with an M500 when you are happy with it, and be sure there is an M501 near the end of your config.g to load the stored calibration.

Rod length is 290.8mm, since they are SeeMeCNC's injection moulded rods, and lots of other people are using them successfully. And I have verified that measurement.

@garyd9 said in Delta - Can't get first layer flat:

Ooooh! Welcome to my world of the past few days.

So, people are going to tell you about adjusting delta radius, but duet should figure that out automatically when you do auto calibration. Then they'll tell you about arm length, but I'm convinced that the duet auto calibration will give you a reasonable flat build plane (via adjustment of the delta radius) regardless of your arm length. Sure, your prints will be dimensionally inaccurate, but your first layer will be good. (@dc42, is that a true statement?)

However, as I've been finding out, the duet auto-calibration makes it's calculations of the delta radius (and other things) based on probing. If the probing results give bad data, you'll get bad data out.

The very low deviation values you are seeing don't tell you that the data is good - it only tells you that it's consistent. Even if there's no deviation at all, it just suggests that you're getting consistently bad data. It is my current understanding that in most (but not all) cases, bad probing data that is consistent is the result of something related to probing being misconfigured.

Based on what your experiencing, and that your zprobe mechanism is a set of FSR's under the build plate, I suspect that your problem is either that you haven't checked the z trigger height AT EACH PROBE LOCATION, or that that data needs to be updated. For that information, check here: https://duet3d.dozuki.com/Wiki/Calibrating_a_delta_printer#Section_Adding_trigger_height_corrections_to_the_bed_g_file (You want the section for adding trigger height corrections to the bed.g file.)

Good luck and please update us on how things work out.

One thing that's not very clear is if I should run the trigger heights before or after a few initial calibration runs? Or does it not matter very much?

Right now I am probing on a 130mm circumference on a ~270mm build plate, which is inside the triangle formed by the towers, but I'm not sure if i get better results by probing a larger or smaller circumference.

I'm having trouble getting my Rostock Max V2 to print evenly on the first layer.
I have the following changes from stock
Cast Alum build plate with FSRs set at each tower (FSRs are as close as possible to each tower)
The build plate is not bound by cups, so binding can't happen.
Duet3d Wifi

FIRMWARE_NAME: RepRapFirmware for Duet 2 WiFi/Ethernet FIRMWARE_VERSION: 2.01(RTOS) ELECTRONICS: Duet WiFi 1.0 or 1.01 FIRMWARE_DATE: 2018-07-26b2

Dedicated 12V PSU, 500W 120VAC heated bed with SSR and safety relay.
Arms and sockets are the injection moulded type, carriages are melamine (I found these to have less play than the injection moulded ones).
Changing back to the melamine carriages from the injection moulded ones improved my results.

I've gone through the Calibration guide for a delta printer.
The towers are parallel with each other within 0.25 mm from the top to the bottom
The towers are 90° to the build plate (checked with framing square)
Mechanically everything is sound according to the guide.

One thing that is not clear to me in the guide, is the Homing guide here: https://duet3d.dozuki.com/Wiki/Calibrating_a_delta_printer#Section_Homing_the_machine
It states to disable the move down from the end stops. Is that supposed to stay disabled all the time? I've assumed that it does, but I'm not sure on it.

These are the results for the Z probe repeatability:

6:33:50 PM G32 bed probe heights: -0.009 -0.009 -0.009 -0.009 -0.009 -0.009 -0.022 -0.009 0.003 -0.009, mean -0.009, deviation from mean 0.006
6:33:42 PM G32 bed probe heights: 0.039 0.039 0.039 0.039 0.039 0.027 0.027 0.027 0.014 0.039, mean 0.033, deviation from mean 0.008
6:33:34 PM G32 bed probe heights: -0.018 -0.030 -0.043 -0.043 -0.043 -0.018 -0.030 -0.043 -0.043 -0.030, mean -0.034, deviation from mean 0.010
6:33:27 PM G32 bed probe heights: -0.011 -0.011 -0.036 -0.011 -0.011 -0.023 -0.011 -0.036 -0.036 -0.023, mean -0.021, deviation from mean 0.011
6:33:19 PM G32 bed probe heights: -0.024 -0.036 -0.036 -0.036 -0.036 -0.049 -0.036 -0.036 -0.036 -0.049, mean -0.038, deviation from mean 0.007
6:33:11 PM G32 bed probe heights: 0.002 0.002 -0.011 0.014 -0.011 0.002 0.002 0.002 0.014 0.002, mean 0.002, deviation from mean 0.008
6:33:04 PM G32 bed probe heights: 0.026 0.039 0.014 0.014 0.039 0.026 0.014 0.051 0.014 0.026, mean 0.026, deviation from mean 0.012
6:32:56 PM G32 bed probe heights: 0.050 0.050 0.025 0.037 0.050 0.050 0.050 0.037 0.050 0.037, mean 0.044, deviation from mean 0.008
6:32:48 PM G32 bed probe heights: -0.058 -0.033 -0.058 -0.058 -0.045 -0.058 -0.058 -0.058 -0.058 -0.070, mean -0.055, deviation from mean 0.009
6:32:40 PM G32 bed probe heights: -0.039 -0.052 -0.064 -0.064 -0.077 -0.077 -0.064 -0.052 -0.064 -0.064, mean -0.062, deviation from mean 0.011
6:32:32 PM G32 bed probe heights: -0.038 -0.063 -0.076 -0.076 -0.063 -0.051 -0.051 -0.038 -0.063 -0.063, mean -0.058, deviation from mean 0.013
6:32:24 PM G32 bed probe heights: -0.026 -0.026 -0.038 -0.038 -0.026 -0.013 -0.038 -0.038 -0.026 -0.038, mean -0.031, deviation from mean 0.008
6:32:16 PM G32 bed probe heights: -0.066 -0.079 -0.066 -0.079 -0.079 -0.079 -0.066 -0.066 -0.079 -0.066, mean -0.073, deviation from mean 0.006
6:32:08 PM G32 bed probe heights: -0.020 -0.033 -0.045 -0.033 -0.033 -0.033 -0.033 -0.020 -0.045 -0.020, mean -0.032, deviation from mean 0.009
6:32:01 PM G32 bed probe heights: -0.013 -0.013 -0.013 -0.038 -0.000 -0.000 -0.013 -0.013 -0.025 -0.013, mean -0.014, deviation from mean 0.010
6:31:53 PM G32 bed probe heights: 0.009 0.034 0.009 0.021 0.009 0.009 0.021 0.021 0.021 0.009, mean 0.016, deviation from mean 0.008

I've tried 4, 6 and 8 factor calibration, but I get similar results with each.

G32
Calibrated 4 factors using 16 points, deviation before 0.026 after 0.021
Diagonal 290.800, delta radius 140.313, homed height 671.777, bed radius 140.0, X 0.000°, Y 0.000°, Z 0.000°
Endstop adjustments X0.65 Y-1.15 Z0.50, tilt X0.00% Y0.00%

If I increase the radius of the bed.g file, the deviation goes up, and if I decrease it it goes down.
This the heightmap:

heightmap.csv
I'm not really sure how to get the heightmap to be flatter.
And if I print a one layer calibration object, (printed without mesh compensation)
Picture of calibration circles
Near the Z tower the nozzle is too close to the bed, as well as near the Y and X towers, so similar to the heightmap results.

What other things can I try to improve the 'flatness' of my bed?

Bed, config, and config override files here:
bed.g
config-override.g
config.g