Klipper Qidi X-CF Pro

Why

The <prior upgrades> to my X-CF Pro work very well.  The part quality was exceptional.  The convenience - top notch.  Why not make it go faster.  I had no idea the neXTG Fiber from Drop Effect would be so fantastic when I started down the road.  It has zero stringing with 0.5mm retractions.  The machine now prints at:

24 mm3/s flowrate (previously 5 mm3/s)

400 mm/s velocity (previously 50 mm/s)

9000 mm/s2 acceleration (previously 1000 mm/s2)

Where a benchy used to print in 95 minutes, it now prints in 35 minutes at similar quality level. Print times in general are cut by half, sometimes even better than that. 

You can still print with PLA/TPU/PET/ABS/Nylon/PAHT-CF/PET-CF/ABS-GF25 etc. all from this single setup.  But now you can do it in at least half the time.  On some models, within one third of the time. 

Orbiter V2 - $45 USD [re-used my prior]

nEXT  G Fiber by Drop Effect - $99 USD

4020 Blower Fan 24V - $7 USD each, quantity 2

BTT Manta 8p V2 - $60 USD

CB1 eMMC (1GB + 32 GB) - $33 USD 

TMC 2209 V1.3 - $18 USD for QTY 5

BTT Relay V1.2 - $15 USD

2510 Axial Fan 5V - $9 USD, three-wire protects against heat creep

Orbitool 02 tool head board - $46 USD, cannot use the included USB cable due to strain relief and short length

USB Umbilical cable -  $17 USD, you need the 2.15m length & no strain relief

Igus Cable Chain - $20 USD, you need 300 mm length

5" HDMI Screen - $43 USD

HDMI to micro-HDMI cable - $8 USD, you need at least 4ft for routing

USB A to micro-USB cable - $7 USD, you need at least 4ft for routing

Bearings - $24 USD, you need QTY 6 if replacing OEM plastic ones

Idler Pulleys - $6 USD, you need QTY 3 with preferred 16 teeth and 6mm wide, but 20 teeth also works if replacing OEM plastic ones, careful to setup Klipper correctly for rotation distance if your tooth count doesn't match mine. 

Drive Pulleys - $6 USD, you need QTY 3 with preferred 16 teeth and 6mm wide, but 20 teeth also works if replacing OEM plastic ones, careful to set up Klipper correctly for rotation distance if your tooth count doesn't match mine.

Tools Used

1. 3D Printer: Print all shims & spacers in ABS or PA12 or equivalent high temperature resistant and stiff material.  
2. Soldering Iron: I spliced the USB umbilical to pass it through the IGUS chain since the XT60 connectors were too large to fit.
3. Dremel: Cutting and trimming various parts
4. Drill: M2 drill bit 
5. Wire Crimps

Step 1

Remove material from the aluminum block where the BLTouch will be mounted.  Since the neXTG hotend is shorter than Rapido V2 we need to ensure sufficient clearance.

Step 2

Install a 12.8 mm length teflon tube with OD 4mm & ID 2.0 mm into the throat of the hotend.  It should reach the bottom side of the extruder as seen in this cross-section. 

Step 3 

Re-using the prior Orbiter that has the brass nut pre-installed, attach extruder/hotend/BLTouch.

Step 4 

Attach Orbitool 02.  Crimp all wires to be compatible with PCB XH 2.54 connectors. 

Step 5 

Remove the surplus plated linear rail that is attached to the X-axis rail to make space for the cable chain.  Install the cable chain with a single M3 screw on left side. 

Mechanical Assembly 

I experienced a problem that took me a while to sort out if it was my own learning curve with Klipper, or if I had poorly assembled something, or if something needed to be upgraded to higher quality.  Long story short, I think all that needs to be done is removing these tension springs from each belt if you want to avoid vertical banding like shown below.

I would see this artifact in my prints only when driving X & Y simultaneously.  Pure X move, pure Y move, no banding.  Diagonal 45° move - awful banding. 

I tried upgrading the bearings and pulleys on all axes to all metal parts with high quality tolerances instead of the plastic parts Qidi shipped originally.  The runout on those parts was so bad you could see the wobble by eye during rotations, no measurement equipment necessary.

There was some improvement as seen below but banding was still noticeable at the speeds I wanted to move at.

Next, I tried removing the mechanical coupling rod that runs across the front top of the machine which enables the single Y-axis motor to drive both the left and the right belts simultaneously.  I figured I had a spare motor left over from another printer and the M8P board had available motor drive ports and Klipper has the ability to dual drive any axis when configuring the firmware correctly. I re-used a stepper motor driver for the Y2 motor, and used the Orbitool's on-board driver for the extruder so I wouldn't have to buy another stepper driver.  All I needed was a mount to hold the motor to the frame.  Use ABS if you decide to do the same.  PAHT will deform from the heat, ABS+GF will crack over time from the mechanical loads.  The gray part is symmetric.  You will need one for each front motor if performing this part of the upgrade.  It is just some M3 hardware and ensuring no belt rubbing on pulleys. I'm not convinced it was necessary.  But I'm glad it was done. 

This also did not solve the banding problem.  It did increase the torque on the Y-axis significantly and I was able to drive much faster in that direction.  I decided to buy the LDO Speedy Power motor <LDO Motors 42STH48-2504AH> for the X-axis as much for the experiment of how it would play out for the banding as well as to match the torque output of the dual driven y-axis.  The 48 mm body length fit perfectly as a drop-in replacement.  Speed was wild - 600 mm/s.  New zippier sounds were heard.  Banding was still observed.   

Opened up the machine again and removed the tension springs on the belts.  Needed to modify the Y-motor mounts so they could apply more tension on the belt that the removal of the spring created.  

Perfect prints. 

Unfortunately, since I took the long route to this solution my machine is now much more upgraded than is likely necessary.  I have neither the time nor the desire to remove all the upgrades and test just the removal of the tension spring alone with no other change.  But I lay it all out here so you can make your own decisions.  I don't run the machine at crazy speeds because good quality is always a delicate balance of cooling, material selection, part geometry, etc.  I standardized on 400 mm/s travels, 300 mm/s infills, 200 mm/s outer wall, 9k acceleration for normal printing, 4k acceleration outer wall, and 9k acceleration inner wall.  This is the new default, I have Orca profiles that are faster for PLA and ABS or slower for fiber filled hybrids.  The speed limitation is always either the hotend with its 24 mm^3/s flow rate, or my derating settings for quality.  I found that when installing the hotend extender that comes with the Next G Fiber I can print at 35 mm^3/s flow rate but my retractions aren't as quick.  Counter to popular recommendations I run retractions on all my jobs with a z hop equal to my layer height.  I find retractions give me much better quality more consistently across geometries and materials.  The print quality is as good or better than with the OEM hardware. 

Electrical Assembly

Below is the basic connection diagram.  

The OEM endstop wires are 4-wire.  M8P ports are three wire.  You can connect them as shown in the photo.  Tie the white and yellow wires together on the same crimp and it goes in the middle. 

I use the 3-wire Delta fan that runs on 5V (Part number: ASB02505SHA-AY6B) that reliably pushes air at a rate of 4.3 cubic feet per minute while matching the height of the cooling fins on the hotend at 25 mm x 25 mm.  Since the Orbitool Toolboard PCB runs on 24V I had to insert my crimped wires into the plastic connectors in a very non-standard way.  Here is how I hijack 5V from some pins that were meant to drive LEDs but still use the tachometer monitoring feature on pin PA8.  Tachometer monitoring guards against hotend melt downs if the fan were to fail.  I only needed that melted plastic mess to ruin my day one time on an earlier printer to learn my lesson.  Now my printer stops if the fan RPM drops below 10,000 during a print.

Software Install

Klipper only allows a single endstop sensor on each axis.  Since Marlin allowed multiple the X-CF pro shipped with two Z-axis micro switches on the floor.  I left both my switches physically connected but configured neither for Klipper.   Klipper has a nice feature to use BLTouch as the sole Z-axis limit switch.  I strongly recommend using it because it will eliminate all variability across the 300mm lead screw distance of the Z-axis.  You don't want to be homing so far away from your nozzle - that is a direct contributor to the variability and first layer adhesion problems with the stock machine. You can reference my setup that gets updated regularly.

<Github Config Backups>

Follow <these instructions> to flash the BTT Relay V1.2 so that it has 30 seconds timeout instead of 10 seconds.  The CB1 takes 30s to boot.  The relay will constantly power everything down otherwise. You can bypass the OEM power button on the top of the machine entirely by simply attaching a jumper between 5V & GND on the relay.  In that way the machine will always receive power and you will need to reach around the back right rear of the machine to toggle the main power switch. 

Acknowledgements

Thanks to tin_tad for the board holder which I did not need to remix.

Thanks to Github user MikeyBigs for the inspiration from his modifications. 

Video

Lately, I've been focused on quality across materials, geometries, and other corner cases.  So, you will see other differences in my machine in the video below. 

1. a 120 mm auxiliary fan for extra cooling when printing faster with PLA
2. enclosure insulation to maintain consistent temperature especially for ABS
3. floor filter for circulation to keep top and bottom of chamber within 5°C for tall prints and to reduce smells
4. temperature sensor to have Klipper dynamically preheat the chamber air temperature with the fans and bed heater
5. inline heater to keep humidity out of my filament
6. three-wire hotend fan so RPM can be monitored and power cut if the fan fails

I saw an easy path to more speed by upgrading the two Y motors to something intentional instead of re-using spare motors.  LDO 2504 motors were placed on the Y-axis.  The ABS motor mounts were replaced with simple garage machined aluminum mounting plates for better heat resistance.  The TMC2209 drivers were replaced with <TMC2240> for the added current it could supply to the 2.5A capable motors as well as their ability to report their own temperature.  Since I intend to drive them hard I want confirmation they are not overheating.  The <TMC Autotune package> was installed to run the drivers more efficiently and generate less heat.  Here is a 16 minute benchy running at 300 mm/s for all internal/external walls, 10k acceleration walls, 20k acceleration travels, 500 mm/s travels, the melt zone extender for the hot end capable of 35mm^3/s flow rate, and many quality de-ratings removed simply to demonstrate that the stamped steel frame and linear rails this machine was built on is pretty good.  A more realistic time for no reduction in quality is 20 minutes.  Simply adding 0.5 mm retractions and 0.2mm Z-hop go a long way to improve quality.  I also added a filament cutter for the BoxTurtle so the toolhead looks a bit different than earlier videos. 

I decided to try a 48V motor supply.  I thought I could get up to 450mm/s print speed at 0.20mm layer height with a 0.40 mm line width to match the 36 mm^3/s volumetric flow rate of the NextG Hotend (450 * 0.40 * 0.20 = 36).  I ordered this <48V PSU> and three <5160 drivers>.   The photos are of a 12 minute benchy at 0.20mm layer height at 350 mm/s external and 450 mm/s internal walls at 25k acceleration for all movements.  48V can go faster but as you can see my cooling capacity can't keep up for benchy overhangs at those speeds.

Fortunately there are a variety of objects to print that are not benchmark models where I can still happily use all the extra speed for fun and learnings.