Upgrading the large delta printer

In my previous entry I described my large Kossel-derivative delta printer. Since building it, I have upgraded it in a number of ways, which I will now describe.

Power supply output grounding

While taking some measurements on my machine, I discovered that the negative side of the output of the 12V LED power supply I am using was not connected to ground inside the power supply as I expected. This presents a safety hazard. So I added an additional wire from ground to the spare negative output terminal of the power supply.

E3Dv6 hot end

My J-head hot end was working very well, until a metal tooth from one of the Bowden tube pneumatic connectors broke off and got carries into the hot end. I managed to remove the tooth, but it seems to have damaged either the PTFE liner or the metal barrel, because after that the hot end kept clogging. So I decided it was time to try a genuine E3Dv6 hot end. The diameter of the end of the top of the barrel on the E3D was the same as for the J-head, so I did not need to modify the effector.

The E3D kit comes with a 30mm fan and a fan duct that clips on the heatsink. However, I needed to mount the IR sensor as well, so I chose to adapt my variant of the J-head fan duct for the E3D, which uses a 40mm x 10mm fan instead. My design for the original version 1.0 IR sensor is here and my design for the version 1.1 sensor is here.

The E3D hot end required a change to the config.g file to account for the different thermistor. The correct B value for the thermistor needs to be entered into the M305 P1 command. E3D says to use a B value of 4267, however from looking at the published resistance vs. temperature data for the thermistor, I believe a value of 4388 will provide more accurate readings at PLA and ABS extrusion temperatures.

The print quality with the E3Dv6 is generally similar to the quality I experienced with the J-head, except that with the e-Sun PLA filaments that I mostly use, the E3D leaves small blobs on the print. I believe the blobs are left where the filament was retracted; so it may be that I will be able to eliminate them by tuning the retraction setting. The E3D Lite permits greater retraction lengths than the all-metal E3Dv6 does, so it may be a better option for machines with long Bowden tubes such as this one.

One thing I particularly like about the E3Dv6 is the plastic collar that holds the Bowden tube. This works like a pneumatic connector in that you push it down to release the tube. However, unlike the pneumatic connector, it doesn’t seem to wear out, neither does it wear the tube.

Verdict: the E3Dv6 is well-made and lightweight enough for a delta. My main reservation is the recommended limit of 2mm retraction, which seems very little for a Bowden extruder. I will be happy with it if I can eliminate the small blobs.

Update: I solved the problem of small blobs by configuring 8mm retraction at 100mm/sec.

0.9 degree/step tower motors

My printer uses 20-tooth pulleys and a 2mm pitch belt. This means that when using standard 1.8 degree stepper motors and 16x microstepping, there are 80 microsteps/mm, giving a theoretical vertical resolution of 12.5um. However, not all microsteps are the same size, and the torque provided when moving from one microstep to the next is low. So the actual vertical resolution will be somewhat less.

Changing to higher microstepping drivers won’t help improve the resolution much because the torque provided to move from one 1/32 microstep to the next is even lower. However, 0.9 degree/step motors now cost only a little more than standard 1.8 degree/step motors, and provide almost as much torque between steps even though the step size is half as much. So switching to 0.9 degree/step motors should provide double the resolution, if the motor resolution is not swamped by backlash or other mechanical factors.

I didn’t have any evidence that the resolution provided by my 1.8 degree/step motors was inadequate; but I haven’t tried printing at layer heights below 0.2mm yet. Nevertheless, an increase in vertical resolution would appear to be useful, especially if I want to print at 0.1mm layer height. Hence my decision to upgrade to 0.9 degree/step motors for the towers.

The original motors were 17HS19-1684S, which are 48mm long Nema 17 motors rated at 1.68A with a holding torque of 0.45Nm. The 0.9 degree/step motors I chose were type JK42HM48-1684 sourced from Etang Electronics via eBay UK. These are also 48mm long motors, rated at 1.68A and with a holding torque of 0.44Nm. I subsequently discovered type 17HM19-1684S, which is the 0.9 degree/step variant of the original motors, with holding torque also 0.44Nm, available here. Shorter motors such as the JK42HM40-1684 or even the JK42HM34-1334 would probably be entirely adequate if you are not too fussy about maximum acceleration.

The motors from Etang Electronics did not have flats on their shafts. Although flats are not essential for securing the pulley, I ground shallow flats on the last 10mm of the shafts anyway. They come with flying leads instead of the JST connectors on the original motors, but too short to reach the Duet, so I spliced them to my original motor leads.

After that, it was a case of making appropriate changes to the config.g file. I doubled the steps/mm in the M92 command from 80 to 160. I soon discovered that the previous maximum speed of 20000 mm/min (333 mm/sec) was not achievable, so I halved the maximum speeds in the M203 command to 10000 mm/min. Increasing the maximum speed will require an increase in supply voltage to 24V and possibly some firmware improvements.

Do the 0.9 degree.step motors make any difference? Yes, but not much. I printed a 50mm spiral vase test cube before (upper image) and after changing the motors (low image). There is a more noticeable moire pattern with the 1.8 degree/step motors. Under magnification, I could see that the layer alignment was also more consistent using the 0.9 degree/step motors.

Verdict: upgrading an existing printer from 1.8 to 0.9 degree/step motors is probably not worth it for the majority of users. On the other hand, if you are buying parts to make a new delta printer, the additional cost of 0.9 degree/step motors is small and worth paying if you are not on a very tight budget. But I suggest going for a 24V power supply at the same time – which means you also need a 24V heater for the hot end, and a 24V hot end fan (and print cooling fan if you have one).

Extruder design and motor

The original Mini Kossel extruder drive wasn’t perfect. It used the same 48mm high torque motor as the towers, which combined with its gear ratio of about 5:1 meant that the hobbed insert had too much torque. As a result, if the nozzle became temporarily obstructed (e.g. because of over extrusion), the hobbed insert would grind a crescent into the filament, and traction would be lost. It is better for the extruder motor to skip steps when the nozzle is obstructed, so that filament feeding resumes when the nozzle obstruction is removed.

To achieve this, I reduced the extruder motor current to 500mA. Unfortunately, this led to a different problem. The 48mm long motor also has high inertia, which meant that reducing the torque also reduced the available acceleration, to the point where the Load Filament function on the web interface did not work reliably.

To resolve this, I decided to replace the 48mm motor by a 34mm one. I would achieve the same torque as before by running it closer to its rated current, but the lower inertia would give better acceleration. I chose a JK42HS34-1334 motor.

Before I had a chance to fit the new motor, the extruder drive jammed and the large gear on the extruder drive lost its grip on the hex head bolt. On trying to dismantle it to fit a new gear, I discovered that I could not re-use the hex head bolt or hobbed insert because of the superglue that I applied (as instructed) to hold them together. So I decided that I wanted a better extruder design, one that could be dismantled and reassembled.

As a temporary measure, I removed one of the extruder drives from my dual-extrusion Ormerod and used that. I printed a bracket to mount it on the 2020 extrusion. The Ormerod extruder uses a 34mm JK42HS34-1334 motor (the same one that I had already chosen) with a gear ratio close to 3:1. I find it reliable, although filament left in the printer overnight tends to break in the curved inlet path. I use the spring-loaded variant by masonstonehenge and a remix of the herringbone gears by iamburney.

I plan to adapt this extruder design to better fit the Kossel.

RobotDigg metal carriages with belt tension adjusters

At the time that I purchased the metal frame corners from RobotDigg, they also sold metal Kossel carriages, however they were designed for continuous loop belts. Now they also sell alternative carriages which include locking teeth for securing the ends of cut-to-length belts, and a belt tension adjustment. I decided to try them.

Fitting them was straightforward, except that they need different lengths of M3 screws to secure them to the trucks and to attach the rod joints than the printed carriages did. Also, the diagonal rod attachment points are half way up the carriage instead of at the top. This lost me 20mm of print height. However, as I was no longer using the upper triangular frame for belt tension adjustment, I was able to put the top triangle right at the top of the vertical extrusions and put the endstop switch mounts right up against the triangle, which recovered about 7mm of build height.

Verdict: these metal carriages are a delight to use, in particular the belt tension adjustment is more precise.

Coming next: 24V power and electronics under the bed

Locating the electronics on the Z tower made it easy to make wiring changes, such as when changing from the mechanical Z probe to the IR probe. But it looks out-of-place, and it makes the printer harder to transport. I believe there is sufficient space under the bed to accommodate the Duet electronics as well as the PSU, SSR and mains wiring, with safe separation between the electronics and the mains wiring. More details to follow in a future entry in this blog.