Talk me out of (or into) a Delta

Hey all,

I have decided to delve into the world of 3D printing. I was previously waiting for the Wally from Nicolas Seward and was signed up for the beta but that never seemed to materialise. After having had a good look around I feel a Large Delta printer fits the bill for me, I think I understand the pros and cons and don't mind a but of tinkering if I get the results I want. My main criteria in order of importance are cost, print size, print quality, print speed, ease of use and printer size. This seems to point to a delta to me but I am willing to be be persuaded if it can be argued that I can get the same print size / print speed from a similar priced non-delta printer.

I am in the UK and I have been looking at the Kossel XL from [builda3dprinter.eu] though I have also been also looking at a self build following this blog.

I really don't want to spend more than £400 but would if there was a real benefit.

I would like something as future proof as possible so the ability to add multi colour printing and print using other materials like nylon would be a bonus.

I would say it is only about one thing: Do you want to print quickly or not?

If you want to print quickly then you also need high acceleration especially if your models will have a lot of holes. And if you need high accelerations then you need a delta. And my suspicion is that it needs to be delta with ball angle joints and not magnetic joints. I do not know whether magnetic joints can handle high accelerations. People did not respond to my questions about this.

Well and maybe, just maybe, if you want to print very tall objects then delta can make a sense too. Kind of.

Otherwise you want a cartesian printer. If you are OK with a small then P3Steel looks like a best option. If you want something bigger then some version of CoreXY looks best to me. Cartesians are much easier to calibrate, more space efficient, do not need so much CPU resources (you can go for a cheap electronics), typically have bowden-less extruder (so no bowden related problems with precise volume of material to extrude). The only problem of cartesians is that they will not be able to achieve so high accelerations as deltas can do.

Quote
MightyMouth
Hey all,

I have decided to delve into the world of 3D printing. I was previously waiting for the Wally from Nicolas Seward and was signed up for the beta but that never seemed to materialise. After having had a good look around I feel a Large Delta printer fits the bill for me, I think I understand the pros and cons and don't mind a but of tinkering if I get the results I want. My main criteria in order of importance are cost, print size, print quality, print speed, ease of use and printer size. This seems to point to a delta to me but I am willing to be be persuaded if it can be argued that I can get the same print size / print speed from a similar priced non-delta printer.

I am in the UK and I have been looking at the Kossel XL from [builda3dprinter.eu] though I have also been also looking at a self build following this blog.

I really don't want to spend more than £400 but would if there was a real benefit.

I would like something as future proof as possible so the ability to add multi colour printing and print using other materials like nylon would be a bonus.

I have both and to be honest my Cartesian printer is easier to use, but less accurate. I prefer my Delta however and if I need highly accurate parts, I use that printer. I hand built my delta from the Folgertech kit and then ended up redesigning and rebuilding after burning up the plastic parts 3 months in. Metal corners cannot be stressed enough for both stability and longevity reasons. As for flexible materials/nylon, you would need to have a legit hot end and not a cheap knockoff ptfe lined tube so be sure of this prior to buying. It's an expensive mistake to end up buying the wrong one. I would recommend the e3d line of multi extrusion nozzles [e3d-online.com] as they are the leading hot end maker for this purpose. The BOM for the printer is listed below, you can take a look over that and the parts that were used if you would like to acquire all the components yourself or just to get a sense of the cost. I actually run the same board as DC42 and utilize his PanelDue in the build as well. So following that guide would be my choice of the two options you presented.



HNG Hackossel BOM

Edited 1 time(s). Last edit at 11/01/2015 03:28PM by thevisad.

Quote
hercek
I would say it is only about one thing: Do you want to print quickly or not?
And my suspicion is that it needs to be delta with ball angle joints and not magnetic joints. I do not know whether magnetic joints can handle high accelerations. People did not respond to my questions about this.

It can, you need to get the heavier duty magnets, most of the ones that are sold are the lighter versions. I think it's the N52 that support the high accelerations.

Quote
thevisad
It can, you need to get the heavier duty magnets, most of the ones that are sold are the lighter versions. I think it's the N52 that support the high accelerations.

Do you have some number how many mm/s² can N52 magnets do?

Quote
hercek

Quote
thevisad
It can, you need to get the heavier duty magnets, most of the ones that are sold are the lighter versions. I think it's the N52 that support the high accelerations.

Do you have some number how many mm/s² can N52 magnets do?

There is a conversation in the seemecnc forums regarding that, showing that the n52 (the standard magnet) could lift more then the n42, I do not recall anything about a particular mm/s though. Last I read, the n42 couldn't handle speeding up any faster, but when he switched out to the n52 he wasn't able to cause them to fail.

Quote
thevisad
Metal corners cannot be stressed enough for both stability and longevity reasons.

Yep. I would say one should go delta only when speed is important. And in such a case it is important to get:

• stiff frame, it cannot wobble when you push it with side forces of about 50 N
• no noticable play in the diagonal rods, pulleys, ...
• stiff belts (get a steel core if you can)
• hotends able to do high plastic flow
• geared extruder - ratio somewhere in the rage of 1 : 2 or 1:3 ... retract speed is important too (or Nema23 motor for extruder if you want direct one)
• make sure the towers are as perpendicular to the bed as possible; the rest (tower positions and endstops) can be calibrated so it is not that important

Quote
thevisad
There is a conversation in the seemecnc forums regarding that, showing that the n52 (the standard magnet) could lift more then the n42, I do not recall anything about a particular mm/s though. Last I read, the n42 couldn't handle speeding up any faster, but when he switched out to the n52 he wasn't able to cause them to fail.

OK, that would mean that their steppers started to skip sooner before the magnetic joitns started to rattle. Would be good to know what steppers and at what current.

Still I would like to hear some example numbers for maximum accelerations achieveable with magnetic joints. Jerk and accelerations are important ... not so much the speed itself. What is quick is too subjective without the numbers.

The point is that Nema17 steppers can exert about 50N force at the arms at the worst case scenario. There are two arms at one tower. So for magnetic joints to be comfortably in the good enough zone one magnet must be able to hold at least about 25N of force. I do not know how well a N52 magnet holds. And will it not be too heavy when it holds well?

25N of force is no problem for ball joints. Of course one must select joints which do not have play. If the joints are not sold completed (like traxxas) then one must be very sure to put them together properly (there rare some instructions about hot/cold water bath and who knows what). I just bought joints which are sold completed.

Edited 1 time(s). Last edit at 11/01/2015 03:28PM by hercek.

Quote
hercek
I would say it is only about one thing: Do you want to print quickly or not?

If you want to print quickly then you also need high acceleration especially if your models will have a lot of holes. And if you need high accelerations then you need a delta. And my suspicion is that it needs to be delta with ball angle joints and not magnetic joints. I do not know whether magnetic joints can handle high accelerations. People did not respond to my questions about this.

Well and maybe, just maybe, if you want to print very tall objects then delta can make a sense too. Kind of.

Otherwise you want a cartesian printer. If you are OK with a small then P3Steel looks like a best option. If you want something bigger then some version of CoreXY looks best to me. Cartesians are much easier to calibrate, more space efficient, do not need so much CPU resources (you can go for a cheap electronics), typically have bowden-less extruder (so no bowden related problems with precise volume of material to extrude). The only problem of cartesians is that they will not be able to achieve so high accelerations as deltas can do.

I prefer delta printers to Cartesians. I have great respect for hercek because it was his wxMaxima script that inspired me to develop my least squares delta auto calibration routine. However, I disagree with two of the points he makes:

1. A delta printer is easy to calibrate if it runs RepRapFirnware and you have a working Z probe. You can do it with three iterations of bed probing, taking less than 30 seconds each, followed by copying the resulting parameters into the configuration file. Total time less than 10 minutes.

2. Cartesian printers are more space-efficient in height, but delta printers are more space-efficient in desk area. The smaller desk space taken by a delta printer is IMO a big advantage over the Cartesian printer, unless you want such a large build area that the height becomes a problem.

Here is some additional advice I have to offer:

1. For a given layer height, the larger a delta printer is, the greater accuracy you need when building it. Auto calibration (or manual calibration) cannot correct perfectly for leaning towers, or the rods not being quite the same length, or the rod spacing not being the same at both ends. My delta now has a print area 300mm in diameter. I would not want to go larger than that at 0.2mm layer height or lower until I have more experience.

2. CoreXY and HBot designs are the most space-efficient of all, so are probably best for really large printers.

3. For £400 including VAT, you won't be able to buy a good kit for a large printer. You can buy a cheap Chinese kit for that, but you will spend more money upgrading it to get it to print well. Or you could buy a small Cartesian such as the RepRapPro Huxley or small Delta such as RepRapPro Fisher. If you can afford more, talk to Think3DPrint3D about their large delta beta program, which is based on my large Kossel design (see the link in my signature). Or buy their Mini Kossel rev. 3 kit and convert it to a large Kossel later.

Edited 2 time(s). Last edit at 11/01/2015 03:39PM by dc42.

---

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
hercek

Quote
thevisad
There is a conversation in the seemecnc forums regarding that, showing that the n52 (the standard magnet) could lift more then the n42, I do not recall anything about a particular mm/s though. Last I read, the n42 couldn't handle speeding up any faster, but when he switched out to the n52 he wasn't able to cause them to fail.

OK, that would mean that their steppers started to skip sooner before the magnetic joitns started to rattle. Would be good to know what steppers and at what current.

Still I would like to hear some example numbers for maximum accelerations achieveable with magnetic joints. Jerk and accelerations are important ... not so much the speed itself. What is quick is too subjective without the numbers.

The point is that Nema17 steppers can exert about 50N force at the arms at the worst case scenario. There are two arms at one tower. So for magnetic joints to be comfortably in the good enough zone one magnet must be able to hold at least about 25N of force. I do not know how well a N52 magnet holds. And will it not be too heavy when it holds well?

25N of force is no problem for ball joints. Of course one must select joints which do not have play. If the joints are not sold completed (like traxxas) then one must be very sure to put them together properly (there rare some instructions about hot/cold water bath and who knows what). I just bought joints which are sold completed.

I have no clue on magnetics, something I bought into and never implemented due to concerns about this. I bought the N52, but they are sitting in a parts drawer and I never printed the parts. Spec sheet here makes little sense to me. [www.kjmagnetics.com] I ended up buying into a bearing from igus that has less play then the traxxas units.

This guy gives mm/s here [www.youtube.com] he made a version of the Berrybot and quotes outer 100, infil 180, travels 350 mm/s

As for as the space advantage, ... yeah, delta is pretty good if printer tallness is not counted.

And also true, dc42 is right that you may dismiss calibration disadvantage of delta. If we would calibrate even for skewed towers then it may be rightfully claimed that delta is easier to calibrate than cartesian. But as far as I know nobody does this jet for a linear delta 3d printer. Conversion from cartesian coordinates to delta coordinates is not computationally hard even when one calibrates for skewed towers but as far as I know there is no quick to use calibration software support for this case. Note that there are developed generic calibration routines for 6 DOF parallel robots and our linear deltas are a subset. The generic solution requires somewhere around 60 measurements (IIRC). Imprecisely built deltas which are otherwise solid should require significantly less. In the worst case, one can go and simplify the math created for the generic case. Even if the calibration for skewed towers would be too expensive to do on the controller board, it can be offloaded to PC. Otherwise solid delta needs calibration only after each hardware change. It would not matter much if it takes longer.

Maybe the biggest reason why I would say a cartesian is easier to calibrate is that when you look at a bad print from a mis-calibrated cartesian then it is quite obvious what needs to be changed. When you look at a bad print from a delta then you are out of luck if you did not study the impact of the different calibration errors on the printer output.

But if you do not care about how delta calibration works or what are consequences of bad delta calibration then just use dc42's firmware and you are done (if your towers are not skewed )

And also there is that small detail that many things can be meant by "delta calibration":

• original manual endstop adjustment only (I hope this sad period is behind us now)
• original Johann C. Rocholl's segmented linear approximation to achieve leveled bed (maybe not only for Marlin???; this does not really fix any dimensional errors except in the Z direction; good only for people who do not care about proper dimensions)
• then you find that some firmware advertise calibration/autoleveling but it works well only for cartesian and on delta it at best rotates/moves cartesian coordinates (may be this is not true any more but was in past when I was selecting the firmware for me)
• Rich Cattell's callibration (only Marlin; this uses some kind of heuristic to guess the right calibration parameters; it kind of simulates manual calibration in software; may not converge but at least if it succeeds then the printed part dimensions should be right, hopefully)
• Dejay Rezme has some offline support (i.e. not embedded in the firmware) for the proper calibration (only single point probe required in firmware; this uses the right math for the problem; you will get true parts if it converges)
• my maxima notebook for proper tower postion, diagonal rod, and endstop callibration (only single point probe required in firmware; this is only for people who can handle a bit of math; but works good and the parts will be true if it converges)
• dc42's RepRapFirnware (well only RepRapFirmware, but leads to true parts if it converges)

Now notice that small little detail. The methods either do not lead to parts with true dimensions (i.e. the parts can be distorted) or they have that pesky little tag "if it converges". For all the methods it probably means the same:

• it will converge properly if your bed is big enough compared to diagonal rod length.
• it will converge properly if you do not optimize both diagonal rod length and delta radius

At least my (and I think also Dejay Rezme's) method can use hundreds of measurement points, which is rather useful to average out measurement errors.

And the final nail: If at least one of your towers is skewed then you are screwed anyway. Good luck figuring out why your first layer does not stick across whole bed as it should without extensive study of the error functions describing z-error from the skewed towers. But I must admit that skewed towers is probably a very rare problem in the printers built by the community. Well, I actually do not know. Still I would say if you just use dc42's ReprapFirmware (i.e. you must select first the properly supported controller board) then you have a high chance to calibrate easily like a spring breeze.

thevisad: Those numbers does not tell me much at the first look. I need to compute the forces the typical magnetic joint can hold. Maybe I'll do it, maybe I will not. Also notice the maximum acceleration is more important than speed for this.

About the magnets:
A datasheet of the magnets won't tell much, because the ball the magnet attaches to also has great influence.

Bigger ball diameter holds better. I see a big advantage in effectors with only three balls.
Material of the ball has a big influence too.
Smooth surface is important.

Magnets with radial magnetic orientation instead of the common axial orientation would be best in terms of magnetic flux.
But I haven't heard of such, yet.
-Olaf

Quote
o_lampe
Magnets with radial magnetic orientation instead of the common axial orientation would be best in terms of magnetic flux.
But I haven't heard of such, yet.
-Olaf

Isn't it better then to place the magnets in the rods and the balls on the effector?
Like Haydns solution.

Quote
dc42

Quote
hercek
I would say it is only about one thing: Do you want to print quickly or not?

If you want to print quickly then you also need high acceleration especially if your models will have a lot of holes. And if you need high accelerations then you need a delta. And my suspicion is that it needs to be delta with ball angle joints and not magnetic joints. I do not know whether magnetic joints can handle high accelerations. People did not respond to my questions about this.

Well and maybe, just maybe, if you want to print very tall objects then delta can make a sense too. Kind of.

Otherwise you want a cartesian printer. If you are OK with a small then P3Steel looks like a best option. If you want something bigger then some version of CoreXY looks best to me. Cartesians are much easier to calibrate, more space efficient, do not need so much CPU resources (you can go for a cheap electronics), typically have bowden-less extruder (so no bowden related problems with precise volume of material to extrude). The only problem of cartesians is that they will not be able to achieve so high accelerations as deltas can do.

I prefer delta printers to Cartesians. I have great respect for hercek because it was his wxMaxima script that inspired me to develop my least squares delta auto calibration routine. However, I disagree with two of the points he makes:

1. A delta printer is easy to calibrate if it runs RepRapFirnware and you have a working Z probe. You can do it with three iterations of bed probing, taking less than 30 seconds each, followed by copying the resulting parameters into the configuration file. Total time less than 10 minutes.

2. Cartesian printers are more space-efficient in height, but delta printers are more space-efficient in desk area. The smaller desk space taken by a delta printer is IMO a big advantage over the Cartesian printer, unless you want such a large build area that the height becomes a problem.

Here is some additional advice I have to offer:

1. For a given layer height, the larger a delta printer is, the greater accuracy you need when building it. Auto calibration (or manual calibration) cannot correct perfectly for leaning towers, or the rods not being quite the same length, or the rod spacing not being the same at both ends. My delta now has a print area 300mm in diameter. I would not want to go larger than that at 0.2mm layer height or lower until I have more experience.

2. CoreXY and HBot designs are the most space-efficient of all, so are probably best for really large printers.

3. For £400 including VAT, you won't be able to buy a good kit for a large printer. You can buy a cheap Chinese kit for that, but you will spend more money upgrading it to get it to print well. Or you could buy a small Cartesian such as the RepRapPro Huxley or small Delta such as RepRapPro Fisher. If you can afford more, talk to Think3DPrint3D about their large delta beta program, which is based on my large Kossel design (see the link in my signature). Or buy their Mini Kossel rev. 3 kit and convert it to a large Kossel later.

I would rather spend a bit more now than have to convert a smaller printer later. Those kits at Think3DPrint3D look okay but I have to wonder why they are much more expensive than the Folger units which use metal corners. I suppose it is mostly down to the electronics used? Mega vs Duet+DuePanel, otherwise they seem pretty comparable but with the metal corners on the Folger Tech unit I would be more inclined to buy that and upgrade the electronics when necessary. The print size is also significantly bigger on the Folger Tech unit. You mentioned Think3DPrint3D have a large beta unit, do you know the print area?

I have a spare touch-screen laptop so don't really need the PanelDue, What are the real benefits to this?

I am more concerned with desk area rather than overall printer volume so a delta still wins in this area for me. The speed benefit is a bonus and I plan on building a solid, straight machine so hopefully the calibration hopefully won't be too much of an issue.

Edited 1 time(s). Last edit at 11/02/2015 06:29AM by MightyMouth.

As for the FolgerTech metal corners there's a thread dedicated to bashing the quality of them, also in the FT kossel thread they are mentioned in a not so good fashion.

The Robotdigg metal corners is of much higher quality, there's also some "simple" metal brackets in the works:
[groups.google.com]
I'm planning on ordering a set of those for my slightly larger delta build.

I hadn't heard that about the Folger Tech Metal Corners, I had seen the Robotdigg corner when I was considering self sourcing the parts but that way seem to cost as much or more than just buying a kit. I filled the form in for the simple metal brackets also as that seems like a good option.

I took DC42's advice and contacted Think3dPrint3d.co.uk about a larger delta printer and had an email from Roland and will now be included in the beta testing of the new Big Delta. As I am confident of the quality of the Mini Kossel they supply I don't anticipate any problems with a beta test version as really it should be all the same parts with just up-scaled extrusions and diagonals.

The more I have a look at T3P3 (as dc42 calls them) the happier I am with the possibility of having one of their machines, I have been through the build manual for the Mini Kossel and it is clear and detailed. Having looked at the Folger tech build guide and having read more on the technologies used in each I can see now why the T3P3 printer is more expensive and I am more confident in the build. Not that the Folger Tech printer looks bad, it seems great value, just as a first printer build the T3P3 seems like it will be much better for me.

I don't think you will be disappointed with a kit from T3P3. Regarding the PanelDue, I would love to sell you one, however if you are happy to control the printer using the web interface from a PC, laptop, tablet or smartphone then you don't need it. The vast majority of people with Duet-controlled printers don't have a PanelDue. It's different with RAMPS electronics, because then you don't have the built in SD card socket and web interface. You can always add a PanelDue later if you want.

---

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 don't think you will be disappointed with a kit from T3P3. Regarding the PanelDue, I would love to sell you one, however if you are happy to control the printer using the web interface from a PC, laptop, tablet or smartphone then you don't need it. The vast majority of people with Duet-controlled printers don't have a PanelDue. It's different with RAMPS electronics, because then you don't have the built in SD card socket and web interface. You can always add a PanelDue later if you want.

I picked up a PanelDue from DC42 and it is by far the sexiest screen I have come across so far. It didn't work out of box since I bought a difference LCD then the device was flashed to support, but a quick download and flash resolved the issue without a problem. I would highly recommend one down the line as an upgrade.

Deltas look cooler

I will probably get one eventually, it really depends on space and how and where I use the printer. I have a perfectly good touch screen Dell latitude XT which I never turn on that I can use but with a web interface will I even need that? I am really excited to get going now, I hope its not too long before the printer is available. I have watched the youtube videos of people building the Mini and read the build guide twice. Roland said he should have more info in the next week.

Edited 1 time(s). Last edit at 11/03/2015 04:58PM by MightyMouth.

One point about the web interface is that it is not as reliable as the panel due. The ajax connection can be lost, it doesn't always respond/refresh, and sometimes the server doesnt resolve at all. The paneldue ensures you always have control and readouts from the machine.

Quote
n8bot
One point about the web interface is that it is not as reliable as the panel due. The ajax connection can be lost, it doesn't always respond/refresh, and sometimes the server doesnt resolve at all. The paneldue ensures you always have control and readouts from the machine.

I will see how it goes, I assume I also have the option of connecting a laptop via USB?

Most people find the web interface reliable, only a few have problems with it. n8bot, if the server doesn't always resolve using its netbios name, try giving it a fixed IP address instead. Yes you can use the USB connection instead. These days I only use it for debugging firmware changes.

Edited 1 time(s). Last edit at 11/04/2015 02:35AM by dc42.

---

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
Most people find the web interface reliable, only a few have problems with it. n8bot, if the server doesn't always resolve using its netbios name, try giving it a fixed IP address instead. Yes you can use the USB connection instead. These days I only use it for debugging firmware changes.

Great thanks, I am viewing the PanelDue as nice to have eventually.

While the OP asked the questions I've merely been reading. Looks like a delta is coming this way... The Duet 0.8.5 I ordered was overkill for an I3 and seemed better suited to a Kossel 1.5, so watching all this information being offered pushed me to pop the RAMPS-FD on the Prusa I3 Steel instead of the Duet and order extrusion and corners for the Kossel.

Thanks! My pocketbook didn't need that!

Quote
Viper97

Thanks! My pocketbook didn't need that!

You're Welcome ;D

Just an FYI;

You can run a delta with ramps. Take into account the segment count to do it properly. Maximum speed is also limited, I noticed that the larger it gets the more taxing it is on the CPU.

My personal delta does about 330mm cylinder area by 380mm height. At this size the mega 2560 chokes at anything over 175mm/s and mostly works best at around 100mm/s.... The upside is that I use it to print large objects and as such rarely go under .2mm, I will be experimenting in adopting my hot end to push out .5mm layers at some point with a 1.2mm nozzle.


I sugest you go with a 32bit controller if your going delta. My next delta may very well be a 32bit powered unit.

---

My Personal Blog. Build blog.
[engineerd3d.ddns.net]

Modicum V1 sold on e-bay user jaguarking11

Quote
dc42
Most people find the web interface reliable, only a few have problems with it. n8bot, if the server doesn't always resolve using its netbios name, try giving it a fixed IP address instead. Yes you can use the USB connection instead. These days I only use it for debugging firmware changes.

The problems I'm having could definitely be related to my router, though it is a nice one. A newer Asus RT-N12. I have DCHP disabled, and when the problem appears, using the static local IP does not resolve either. There is definitely nothing responding at the address. Usually the problem is followed by the web interface momentarily losing its settings for the gcode menu dropdown and other features like temp lists. If I do a hard refresh in chrome, it can sometimes resolve itself right away. Sometimes I need to reboot the duet.

Edited 3 time(s). Last edit at 11/05/2015 03:55PM by n8bot.