Starting with a Delta

Hello,

I'm looking into starting 3d printing. And totally subjectively I already decided it has to be a Delta printer. Just because they look more intriguing and more elegant than a Cartesian. Can't really help it, just an engineering curiosity.

So that narrows the search down. But after seeing this tweet of Richrap on a affordable 3DR Delta kit, I wondered which of the following would be recommended.

I already was looking at the builda3dprinter or a mini kossel from think3dprint3d or even a micro delta of reprap-france. And now Richrap points out to the 3DR Delta.

Which one of these would be recommended. I have a preference for the builda3dprinter Kossel XL as it is shipped in the Netherlands( where i live).

But are there other differences to look for?

For instance I question the way the arm are attached with a tie rip in case of the micro and and 3DR.

There is also a choice to be made with the hot end. A v6 or v6 lite.

Looking forward to all the advice!!

I had the Mini Kossel kit from Think3DPrint3D and upgraded it to Duet electronics (also from Think3DPrint3D) as described in my blog. It's a good kit, but the build diameter of 170mm is rather limiting. I've recently upgraded mine to 300mm build diameter. A larger Kossel will be more versatile if you can afford it.

I have no experience of the 3DR or micro delta.

Personally I wouldn't go for any kit that uses Arduino/RAMPS electronics if I could avoid it. That electronics is barely powerful enough to run a delta, a pain to configure (you have to recompile and upload the firmware to make almost any change), tricky to get the motor currents right, has a slow USB connection, a fragile voltage regulator, no SD card, and you can't add a graphics LCD without killing the performance. But if you can only find kits that come with Arduino/RAMPS, I guess you could always upgrade the electronics later as I did.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

I started with a Rostock Mini Pro from 3D Printer Czar. Expensive by today's standards, $459, but free shipping world wide and supposedly no VAT (not a US problem, so don't quote me). Not sure I would recommend a Delta for a first printer due to the calibration process. While easier to build then the alternatives (i3), I think the alternatives will be easier to get working. The build volume is also a somewhat limiting factor for the mini's, so you might research a few projects you would like to print beforehand.

If you do decide to go with a delta though, and they are an elegant option, you will be able to to get by with a cheap 8-bit processor and a small 4-line LCD. I don't recommend the full graphic LCD based on my own personal experience, and you will eventually want to upgrade to a 32-bit processor like the Duet, Smoothie Board, or a smoothie board compatible board. I think the Kossel's will be easier to upgrade to a heated bed thanks to the V-slot design, though in most cases I doubt you will be printing anything hotter than ABS, so the v6lite is probably a good choice over the v6. Keep in mind most kits come with a J-head or all-metal cloned hotend though, which should get you by to start.

With all due respect to RichRap, the 3DR is last year's design. Delta technology has moved on, and even the old 3DR thread here had vanished off the radar for a few months until it came back today for some reason.....

Quote
vreihen
With all due respect to RichRap, the 3DR is last year's design. Delta technology has moved on, and even the old 3DR thread here had vanished off the radar for a few months until it came back today for some reason.....

Ok, but what is today's delta design?

Cherry pi 4 Xl

Building one now.

To sum up these good advise's I conclude the following:

Micro is definitely to small
Xl or other larger ones are recommended because they are more versatile and cost for a larger one aren't that much higher if you start from scratch

The Ramps controller won't cut it and as I understood it is at its end of its developed cycle.
So preferably something with a 32-bit processor like an Azteeg or smoothieboard.

Hot-end advise is if the possibility exists avoid the J-head and go for a V6(lite)

As the Cherry pi is advocated I notice a difference is the V-roler is the newer kind. Am I correct?

I agree with your conclusions, except that the Duet is also a good 32-bit board for running a delta. Currently, it is the only board with firmware that supports fast single-iteration auto calibration (using a least-squares fit), and that supports segmentation-free delta movement.

However, you may not find a kit that has anything other than Arduino/RAMPS, so you may have to make do with that and upgrade later. Think3DPrint3D recently did a variant of the Mini Kossel kit with Duet electronics instead of Arduino/RAMPS for a contact of mine. However, an advantage of starting with a standard kit is that all the wiring looms should be pre-made and the right length, and so easier to put together and get working.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Quote
dc42
I agree with your conclusions, except that the Duet is also a good 32-bit board for running a delta. Currently, it is the only board with firmware that supports fast single-iteration auto calibration (using a least-squares fit), and that supports segmentation-free delta movement.

However, you may not find a kit that has anything other than Arduino/RAMPS, so you may have to make do with that and upgrade later. Think3DPrint3D recently did a variant of the Mini Kossel kit with Duet electronics instead of Arduino/RAMPS for a contact of mine. However, an advantage of starting with a standard kit is that all the wiring looms should be pre-made and the right length, and so easier to put together and get working.

Figuring out wiring won't be a problem. It might be an option to order a kit without a controller and order it separately. At Builda3dprinter I can customize a selection so i could exclude the controller from ordering.

Thanks for all the advice!

Quote
dc42
I agree with your conclusions, except that the Duet is also a good 32-bit board for running a delta. Currently, it is the only board with firmware that supports fast single-iteration auto calibration (using a least-squares fit), and that supports segmentation-free delta movement.

I'm curious what kind of max step generation rate you achieve when doing the segmentation-free thing. I assume the added math has some sort of cost. Very interested in the answer as it'll probably tell us if we want to do the same thing in Smoothie or not.

Tell me please, is there even any difference in print quality between these two things? Segmented vs segmentation-free movements.

The math certainly requires more CPU power than an approach using course segments. Not only do we calculate precise steps timings for long continuous delta movements in the Duet firmware, we also calculate precise step timings during the acceleration and deceleration phases of each movement, on both Cartesian and delta printers. We need to calculate the square root of an integer somewhat larger than 32 bits in the general case, but smaller than 64 bits, once per step in each acceleration or deceleration phase, plus once per step for each delta movement. So that's twice per step during the acceleration and deceleration phases for each movement of a delta printer. I would not have attempted this on an 8-bit processor.

On a delta printer, the worst case appears to be pure Z movement (assuming we don't treat that as a special case - which we don't). My target was to support movement up to 500mm/sec in any direction on the Duet. With 80 steps/mm (typical when using 16x microstepping on a delta), this requires 40K steps/sec for each of the 3 motors - so 120K steps/sec in total.

The processor on the Duet is a Cortex M3 running at 84MHz. The square root function takes about 4.5us to execute. So 6 of them (for 3 motors, times 2) take 27us. Adding on the rest of the step generation code, I found that the shortest interval at which I could calculate and generate the step timings precisely for all 3 motors during pure Z movement, while still responding to the web interface and everything else, was about 50us. This is around double the 25us available between stepper pulses for movement at 500mm/sec.

To achieve 500mm/sec, when the step pulse interval falls below 70us, the code does the calculation less often. Initially it skips one calculation in three, then two in four, and so on. The lack of precise timing doesn't matter at high speeds, because the step pulses are being generated much faster than the stepper driver chips can adjust the current. At the lower speeds used during printing moves, calculation of the step times is always precise. You can think of the code as being equivalent to doing segmentation at the rate of 14000 segments per second.

There is some optimization possible in the code, which I could do but haven't found to be necessary yet, because 500mm/sec is already faster than most users need. Currently I use a 62-bit square root function, but I could probably manage with about 48 bits if I adjust some of the scaling constants. There is also scope for changing the data structures so as to reduce the overheads in the step ISR. This might be useful to do if supporting 32x microstepping.

I can't comment on the difference in print quality compared to using fine segmentation, because I only used a delta printer which used segmentation for a short time. What I can say is that my delta can print at around double the speed of my Cartesian printer without loss of print quality.

Although the non-segmented approach increases the time taken for step generation, it greatly reduces the amount of computation needed in the lookahead system. Apart from possible improvement in print quality due to the smoother movement, it also made it easier to support pre-compensation for elasticity in the Bowden extruder system - although the jury is still out on whether this improves print quality.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].