I was not happy with the first prints out of the delta printer. In spite of upgrading the firmware from Marlin 1.0 to Marlin 1.1.3 and running the auto-calibrate several times, things were not as good as I wanted.

One problem was with one axis end stop, which turned out to be loose on the rail. The square nuts (instead of true “T” nuts) was the problem. After much fussing I was able to tighten it, but I think I’ll be buying some proper “T” nuts soon. With that fixed, things were better, but still not great.

I realized another problem was the “probe offset” might not be correct. The Marlin 1.0 default was 0.2mm, but the Marlin 1.1.3 default was 0.7mm. I was using 0.7 and the first layer height seemed high (it wasn’t printing the first layer at all).

Yesterday (Oct 12) I manually calibrated the printer using the paper strip method, then checked the additional distance from “0” to where the probe end stop triggered. It turns out to have a bit of a dead zone. Using the default test height of 315mm, zero was found at 1.13mm. The probe end stop triggers ON at 0.83 but OFF at 0.93. In the middle is a dead zone (0.83) that is either ON once triggered, or OFF before triggering.

As the auto-calibrate uses the probe trigger point, the distance between zero and trigger on was 1.13 – 0.73 = 0.4mm. This was more than 0.2 and less than 0.7 explaining why the auto-calibrate height number was off.

Using the new depth (315 – 1.13 = 313.87) and probe offset (0.4), The prints should be good.

…

Except they are not.

Watching closely while it started printing, I saw the problem immediately. The bowden extruder mechanism which decouples the extruder from the hot end was not feeding filament properly. Frequently the extruder gear was turning, but no filament was moving. There’s an adjustment wingnut to set the tension, but I found with experimentation that it is just too finicky to work reliably. It seems there is not setting of the tension that will always work first time, and you can’t be fiddling with tension all the time during a long print.

Examining the bowden extruder mechanism, I think I see the problem and the solution. The mechanism is plastic. The idler is a bearing and washers, but the tensioner is just a bolt and wingnut. What is required is a good quality aluminum housing with spring tension similar to that on my cartesian 3d printer that I built.

A quick search on amazon.ca found several excellent choices in the $13 range, so I ordered an appropriate one. It should arrive next week and I’ll resume the testing. I’m quite certain a good extruder will solve the final problems, and perhaps with a final auto-calibrate I’ll be up and running.

Of course, Marlin is now at version 1.1.6, but I’ll hold off trying that until after I get the current version working.

I’ve been busy blowing glass the past week or so, so haven’t spent much time with the 3d printers. I did print 8 more weight collars for the weight bench, installing 4 of them last week. They work really well and are easy to print. Printing 4 at a time is easy and takes about 2 hours. What  is even nicer is the new brand of filament is not jamming at all feeding the original cartesian printer, so I don’t have to monitor the print all the time.

The delta almost printed a nice cube. Only the base was off, and on reflection I think it’s because the setting for ‘probe offset’ is too large. The probe offset is basically the distance between the extruder tip touching the printer base and moving enough (into the base) to trigger the probe microswitch. There were two values in the Marlin firmware – 0.2 and 0.7. I’m using 0.7 right now and I think it’s too much, so when you calibrate the printer the final distances are out by a small amount. Not much, but just enough that the extruder is farther away from the bed on layer one than necessary for a proper first layer.

I need to spend some time with raw gcode to find the actual probe distance for my printer and then use that to recalibrate. Perhaps later this week or sometime next week.

I’ve been working with the FLSUN delta printer I built from a kit this week, and although it has printed a few of the test cubes, it’s still not quite fully functional yet.

It turns out delta printers are a lot faster than cartesian (x,y,z) printers, but with that speed comes some rather unique challenges. The first is that the printer, being tripod and pendulum based, really wants to print in an arc. To counter this, the software needs to know a lot of details about the printer, such as segment lengths and arcs in order to calculate various angles and offsets to use when printing.

Fortunately the delta code firmware I’m using (Marlin 1.1) now has an auto calibrate feature that works with the built-in z-probe on the FLSUN delta. Not all deltas have the z-probe, but it’s easy enough to add to any printer. Basically it’s a microswitch down near the extruder that detects when the extruder is near the print bed. On the FLSUN, the switch is triggered by a rather ingenious lever mechanism built into the extruder mount. When the tip touches the bed, it rocks slightly and trips the microswitch. You can fine tune the triggering pressure so that it’s very accurate. There is one compensating measurement you must take, and that’s how much deflection occurs when triggering. That is, the extruder doesn’t just touch the bed, it touches it a bit MORE than when printing to tip and trigger the switch. You need to know (in mm) how much this extra movement is so the software can compensate when calculating actual bed height.

The auto calibrate touches the bed in a set number of places that can be programmed (n*n, with n=4 as default). It can then calculate various angles and offsets that you then either save in the printer firmware eeprom or hand code into the software. I chose to hand code. The default calibration runs until a set standard deviation is achieved, so it’s quite accurate.

Once the auto calibrate is done and the information encoded in the firmware and uploaded to the printer, you can print.

My first prints were terrible. I quickly realized the extruder calibration was off. Extruder calibration is done with auto calibrate but must be done separately. Mine was barely moving, so after some experiments and adjustments, I got the correct parameter (stepper steps per mm of filament) and that was OK.

However, the next several prints were all still bad. I quickly found one of the three end stops that controls the position of the extruder was loose. After adjusting and tightening, things were much better, but there’s still a problem with the print.

The first few layers (about 1mm or so) are offset from the rest of the print in one direction. It’s consistent and repeatable at this time. The rest of the cube is perfectly straight, so it’s a “first layer” problem. So far I have not found any solution on the internet.

One hint that I’ll be pursuing is that the height measurement of the cube is only 19mm, while the sides are both 20mm perfectly. That implies again a “first layer” problem, possibly too hot. I am also considering the fact that this printer is incredibly fast. It’s almost twice the print speed of my first (cartesian) printer. It’s possible that it is simply printing the first few layers too fast, which may also affect the temperature of those first layers.

My next steps are to compare speed parameters between Marlin 1.1 (my current firmware version) and Marlin 1.0 (supplied with the printer) and my cartesian printer (also Marlin 1.1, but much slower settings). I will also try slowing the first layer in slic3r as a quick test. If slowing the initial print speed works, then that will be great. For now it’s a wonderful printer that has yet to realize all it’s benefits (and yet to make it’s first perfect print).