Page 'Choosing stepper for a delta' created in RepRap Wiki

I created the page 'Choosing stepper for a delta' in the RepRap Wiki :[reprap.org]
with elements from my own experience and informations from forums.

Hi Prz,
It's a good article!

I added a link to an article about the DRV8825 problem.
I added it to the Tutorials category.
I made a few very minor grammar fixes.

I'd like to discuss a few of additional things that are relevant to this discussion:

- The relationship between the choice of pulleys and choice of stepper motor. For instance a worst case scenario is: a .9 degree motor and pulleys that are too small. This requires the motors to spin too fast for rapid moves, and moves where the effector is near the outside of the build area. This can cause loss of steps due to low motor torque at the high speed, and too many steps/second for many 8-bit controllers to handle (causing stuttering). (1)

- Another issue with motor/pulley choice is the desired resolution (steps/mm) - is there a reasonable range that can be recommended? This affects print quality if steps per/mm is too low, however resolution over (x) steps per mm is not advantageous and will limit the speed that can be obtained.

- Mid band resonance. I believe that larger motors (Nema 23) will hit the mid-band resonance regions at lower RPM. This can limit their maximum usable speed unless an external DSP based driver is used. This is my belief and experience, however I can not yet link a source.

- When you speak of "larger motors" are you speaking specifically about Nema 23s?

(1) Smoothie allows setting the maximum speed the cars move on the towers independently of the maximum speed of the effector, using alpha_max_rate, beta_max_rate, gamma_max_rate in the config file. This can eliminate over-speeding the motors when the effector is at the outer edge of the print area and a motor must move at it's fastest.

Thanks for your work.

---

My printer: Raptosaur - Large Format Delta - [www.paulwanamaker.wordpress.com]
Can you answer questions about Calibration, Printing issues, Mechanics? Write it up and improve the Wiki!

I saw the stepper speed problem also in your article about delta geometry. Maybe you can put there chart like this there so that people have better idea how much slowdown (for old Marlin) or stepper speed up (e.g. ro repetier) they get at edges of the print area.
[forums.reprap.org]

As for as the stepper speed up and the moment. Yes one needs higher speed at the edges which leads to smaller torque but on the other side one need less torque at the edges since the stepper is better leveraged when pulling/pushing against platform inertial forces. The required torque will not change as for the the carriage inertia though. So it is worse at edges but not as much worse. Well, probably not important enough to complicate the article.

Edit: And I would argue that it is important to retrace filament quickly (below 0.1s). So if one has geared extruder and a bowden (therefore a rather big retrace length) than extruder may be the reason to go 24V too.

Edited 1 time(s). Last edit at 01/27/2016 01:56PM by hercek.

Hercek,
Good that you do modifications. That a common problem on new articles that people hesitate to work on them. There's no point, as it is the purpose of a wiki.
And for the grammar, for sure corrections shall be done, as english is not my mother language (I'm french).

About the step output difficulties for 8 bit processor and 0.9° motor, I will made the comment, and this was also underlined in another forum (google deltabot)

I don't think there is any point to use Nema23 on a printer , because of the higher inertia, lower microstep precision and maybe, as you tell, mid-band resonances. A note about it may be of interest. Bigger was meant for higher torque, hence longer motors. I have not done a lot of researches about mid-band resonances and chips used in the printer world, but this should be handled by the driver. At least that is what is done by the Gecko for more than 15 years now - with analog electronic !
I have not much ideas about the step by mm criteri, but the range used is thin, as most kossels and derivative are using 16 teeth pulley, and others 20 teeth because that is simpler to supply than 16 teeth.

For side speed (and acceleration, I will set a table, that is easy to do and will help understanding, that just the angle tangent).
Carriage are generally a bit heavy because of the metal bearings, so their inertia may be leading over effector inertia.

I think the acceleration paragraph shall be developed.
And for movement outside the build area, I set up a brass brush on the D-box, which need fairly low angle to be reached. Not only I had to reduce the speed, but also the acceleration for it to work without loosing steps.

For extruder, I made a strange thing on the D-Box, I installed the same 0.9° motor as for movement, because I was not willing to place a specific order only for extruder motor. I had to reduce the jerk as I got troubles with the retract but I use a direct drive and 24V so that is relatively fast anyway. I see no point using geared extruder for 1.75mm filament. I have done it only on HXM to use the low torque supplied motor (the standard RRP 2.2 kg.cm motor).

Paul, about yours corrections,
My aim using the word 'chip' which I agree may may not be the best was to distinguish the integrated circuit from the board. I find the term 'driver' a bit confusing in this regard. Maybe we could let 'driver' for the integrated circuit and write 'driver board' for the complete board ?
As for the link to the article about DRV8825, it was already in the section 'links', and I agree it is really good with excellent scope views.

I apologize to have a bit confused your post with the one of Hercek and mixed stuff.

This depends on the hotend type too but with an old style hotend like this [forums.reprap.org]
I could print at most 60 mm/s without geared extruder. After I added 2:1 gearing the maximum printing speed of 150 mm/s is achievable (maybe more - I did not try). Though I typically print at 120 mm/s because on higher speeds the solid layer over sparse infill takes long to hide the internal infill structure. At higher print speeds than 120 mm/s it takes more than 3 solid layers to completely hide the internal infill pattern. I mean the layer is closed (no holes) but it is visible where the internal infill lines were. I use 7200 mm/s² accelerations and 20 mm/s jerk. At these settings accelerations "waves" are visible after corners but it is not that bad. Too springy belts and too thin rods.
I would like to make a printer which can handle accelerations of about 15000 mm/s². Maybe I'll get to it sometimes

PRZ,
Yes, you could even say "driver chip" and "driver board".
Good, I didn't see the link.

No apologies necessary, thank you for your work!

Note that I certainly do not want to complicate the article.

A bit of background on why I mentioned the three items above:
Not knowing better when I built my delta - I had all those issues. I used .9 degree motors and 17 tooth pulleys, and this seriously limited speed, first due to the 8-bit controller and DRV8825 drivers. When I upgraded to a 32 bit controller the motors hit mid-band resonance with what I considered to be reasonable travel speeds. So I agree that Nema 23's are problematic - but also the combination of my choice in pulleys, controller, and drivers.

I had serious mid band resonance issues for the main steppers until I used DSP drivers and tuned them. The arms on my machine can go to very low angles indeed (I'm well aware that this is not ideal) - the required speedup did cause headaches until it was properly accounted for. When the motors reach a certain speed they go no faster now due to the Smoothie config. This does not affect normal printing, because the cars don't need to go that fast normally. After increasing the pulleys to 30 teeth I still have 177 steps per mm, with .02 mm repeatability, and then could go faster yet. Although I do not recommend Nema 23, this setup is working fine now.

Mid-band resonance occurs fairly abruptly, and is not eliminated by DSP drivers, only reduced - and the motors run much quieter. Using the settings in the Smoothie config I do not have to worry about the speed/acceleration at all now when accessing any part of the bed, it won't go into that speed danger zone.

I used these for all 4 motors on my delta, as the maximum print size is (x/y plane) is large, 430mm, and some of it's parts were designed a bit heavier to allow me to swap out the effector for pick-n-place, laser, drill, etc. I have printed prototypes for someone nearly that large.

Really the motors see hardly any noticeable torque under normal operation compared to what they are rated, it's when they over-speed into low-torque/resonance zone that is the problem - when fully resonant - boom - they have no torque at all. This is a much steeper drop in torque than the RPM/torque curve.

One of those Nema 23 works great for my direct drive 3mm extruder, as their rated output is 11.2 kg/cm. I am running that one at 2A, 26v with a Mk-7 gear bored out to fit, directly from the Smoothie. I don't measure the output by print speed, I use mm3/second extruded - it will comfortably output 12mm3/second for normal layers, 18 when bridging (ABS).

I wish I'd known then what I know now about what that combination of motors, controller, pulleys, and drivers would do. Would have saved a lot of time and money.

So hopefully you understand a bit where I was coming from in my fuzzy headed post earlier.

Edited 1 time(s). Last edit at 01/27/2016 10:05PM by Paul Wanamaker.

---

My printer: Raptosaur - Large Format Delta - [www.paulwanamaker.wordpress.com]
Can you answer questions about Calibration, Printing issues, Mechanics? Write it up and improve the Wiki!

Looks like I cannot do modifications. Forum login does not work for wiki.

Yes, forum and wiki are completely separated sites, so I registered in the wiki with same password and login as in forum.

Paul, your setup is more a CNC setup than a printer one, you could drive a middle size CNC with these, and even a fairly large one if you don't need high speed.
I understand the need for more power, notably for Ceramic printer or multi-purpose machines, so I created the chapter 'I want it big !' [reprap.org]

Sounds good!

Hercek, I was surprised by your comment about the geared extruder, so I done my very first speed test in white PLA (Capifil). I usually use PETG, which extrude very easily, so for me extrusion is never a problem.

To test relatively large extrusion flowrate, I set the layer thickness to 0.3mm with a fixed width of 0.6mm.
I have two delta printers of own design:
- The D-Box, max diameter 286mm, hotend all metal Prometheus dia 0.4 tuned for PETG with three nuts, which makes a fairly long heating zone. Extruder is direct drive, stepper 4.2 kg.cm, hobbed insert MK8. Current 1.5A for a nominal of 1.7 A. That is a good tuning for high speed but uncommon.
- The HXM 'Lily F' max diameter 170mm, all-metal hotend from Fisher delta dia 0.4mm, designed for PLA at low speed, with short heating zone (less than half my Prometheus setup). The extruder is geared with a ratio 3.1:1 on a stepper 2.2 kg.cm, which makes 6.82 kg.cm, something quite equivalent to a 60mm nema17 direct drive, so this is interesting.
Current 1.1A for a nominal of 1.2A

I got a lot of problems of grinding filament with the D-Box extruder [rouzeau.net], I have to set a stronger spring and I am at the maximum possible pressure without slowing down the stepper.
The hobbed insert, supposedly working with both 3mm or 1.75mm filament, is clearly not adapted to 1.75mm.

The HXM geared extruder, [rouzeau.net] with special hobbed insert from RRP, same diameter as an MK8, which have a good grip and with a torque 60 % higher than the D-Box doesn't grind the filament and skip steps. That is a huge difference in quality and I shall supply a better hobbed insert for the D-Box.

I tested first a circular vase diam 160mm on the D-Box (that is 56 % of maximum diameter), chosen because it makes a periphery of 500mm, easy speed calculation.

I was not capable to go over 125mm/s with this setup, and a temperature of 245°C. That is a flowrate of approximately 18~19 mm³/sec, not that bad.

Now, test on the HXM Lily, the speed where I could reliably extrude was not exceeding 80 mm/sec, so a flowrate of 12 mm³/sec and I tried to raise the temperature without noticeable effect. Settled temperature : 255°C
That means with a extruder having a torque 60 % higher, I was not able to get more than 64 % of the flowrate, only because of the heat length difference.

So, you are true, a geared extruder is of interest depending your hotend setup.

My first vase diameter was 150mm on the HXM Lily (that is 88 % of maximum diameter). I got banding effects, while I have none on the D-Box. But thats comparing apples to oranges due to the difference of diameter of the two printers, so I made a new test with a vase of 100mm diameter (59 % of maximum diameter), which is comparable to the test of the D-Box. Quality was much better, with no banding at all. Quality may be slightly better than on the D-Box, but this is at a lower speed (80mm/s instead of 125 mm/s). I have yet to test a large diameter part on the D-Box, to see if I experience the same banding, knowing that the stability of D-Box effector and overall stiffness is much better than on HXM Lily.

Ok, I print almost exclusively ABS (Ø1.75 mm) and always at 230°C. The bowden length is about 60 cm.
If I recall correctly I did all the speed tests at 0.3 mm layer height. But maybe not. It is possible the tests were done at 0.2 mm.
But in normal circumstances I print almost always at 0.2 mm layer height. So I'm sure I can reliably print at 60 mm/s at 0.2 height without gear and 150 mm/s at 0.2 height.

You have seen the picture of the extruder. Ø 0.5 mm nozzle. I do not know the extruder name. From the web posts, it looks like most people use more modern extruders now. Maybe my extruder is not that good and that could explain some differences.

Compared to your D-box:

Torque:
My rostock: stepper is rated at 4.5 kg.cm at 1.68A; but it gets 1.45A ⇒ 3.9 kg.cm
D-box: 4.2/1.7*1.5 = 3.7 kg.cm
Lily-F: 6.82/1.2*1.1 = 6.25 kg.cm
So my stepper is about as strong as your D-box.

Pulley:
My rostock: Ø at the filament touch point is 10.45 mm
Your MK8: 7.1 mm
I have 47% (1-10.45/7.1) disadvantage here. This probably can explain some differences of the differences.

Flow:
My rostock:
* without gear: 60*0.5*0.2 = 6 mm³/s
* with 2:1 down-gearing: 150*0.5*0.2 = 15 mm³/s
D-box: 18.5 mm³/s
Lily-F: 12 mm³/s (interesting should be more, you have stronger stepper there ... but different hotend)

If I would compensate for a different pulley then my numbers would be:
* no gear: 8.8 mm³/s
* with 2:1 gearing: 22 mm³/s

It is possible that my maximum flow rates are 50% bigger (If i made the tests at 0.3 mm and not 0.2 mm).

Here is the extruder I use:

And its OpenSCAD source code: [www.hck.sk]

In general, my experience is that down-gearing the speed 2:1 is best. One cannot go much more (e.g. 3:1) if bowden is used because then it would be hard to retrace reliably in 0.1 s (even with 24V I use). And without quick retrace there may be blobs of plastic at the end of paths.

To see what are the differences between the Fisher and Prometheus hotend, which explains the easier flowability on the Prometheus:



The tape hide a bit the stuff, and what you cannot see is that there is a real heatbreak on the Fisher, while the heatbreak on the prometheus is only the M4 thread, so the real heating length is even higher than the geometrical one. We may have a 1:3 ratio here.
In theory, long heat zone favor stringing, and indeed the Prometheus (with this setup) does have more stringing.
However, as this longer heat zone allow a lower temperature while going at relative speed, this is not as bad as you may expect. With The Fisher hotend, I have to use quite high temperature with PETG and its make it sensitive to speed.