Carbon rods: 6mm solid or hollow 10mm dia?

The headline tells it all.
With Haydns Delrin Mag-arm-kits, I have the option to use up to 10mm dia carbon rods, but the 360mm rods I got from T3P3 are 6mm ( hollow or solid, I can't tell )
Although, I get the best print results so far on a Delta, I see ringing.
DjDemonD uses 6mm solid, but my gut feeling wants to use 10mm hollow rods instead.

Will I see significant disadvantages with 10mm rods? eg, lower printable radius?

I don't envisage any issues with using 10mm rods instead of 6mm. But the ringing you observe may not be caused by the rods flexing, it may be caused by the belts.

I too would prefer 10mm hollow rods over 6mm solid.

Edited 1 time(s). Last edit at 02/21/2018 07:43AM 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].

Diagonal rods are loaded mostly in the direction of their axe. If the ball joints move freely it is unlikely that rods would cause ringing. I use Ø 5 mm rods and they seem to be strong enough.
Belts, motors (small static moment at higher microstepping), or smooth rods are more likely cause of ringing.

My XL has rods that are 360mm and are Haydns pre-made delrin ones with 6mm rods. I was a bit dubious that they would be stiff enough but I'm not sure I could make that machine more accurate (did I really just say that about a 3D printer?).

But I do need to make some new rods for the fishtank printer in abs or petg as it gets quite hot inside, if I can sort the leadscrews out first. So I could try the 10mm carbon tubes (assuming they fit over the Haydn rod ends?) and then I could swap rods and do some fairly well controlled testing.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

Solid or hollow: If you have two rods with the same mass and length then the hollow rod will be stronger than solid one. Prefer rods with wall thickness around 0.5, at most 1 mm.

Thanks everyone for their input.
I made a ringing test with both variants at 120mm/s perimeter speed and they both resulted the same, so it's the belts fault.
But the 10mm rods look much better

That's a pity, since you want to print bigger parts with a bigger printer and have to go slower, because the belts are causing issues.
Just a shoot from the hip: Why not place the idlers on separate brackets ( not the top corners ) and adjust belt length according to the current part height. That way, we'd always have the lowest ringing possible...

@DJ 10mm rods fit into the rod ends not over.

Tell us more about the belts? Are they stretching?

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

I was thinking about a design using Ø 1.6 mm steel cable instead of belts to minimize ringing. Drums would be used to drive the cable (similar like spectra string kosel). Ø 16 mm smooth rods instead of linear rails.
I probably never get to build this but here are some estimates of different error kinds for such a printer when using Nema17 motor (42 Ncm):

• target acceleration: 7.3 m/s²
• microstep length: 14 µm
• static positioning error due to drum length: 14 µm (does not exist with belts)
• stepper motor dynamic error (because static moment is limited): 14 µm
• dynamic positioning error due to cable elasticity: 7 µm
• maximum dynamic smooth rod bending error: 5 µm

The dynamic errors are the errors which cause ringing.
The belts are typically the biggest problem. If the above printer would use 6 mm wide steel core belts then the belt stretch error would be around 127 µm. If it would use Gates glass core belts then it would be about 2 times bigger (254 µm). And this is still only 7.3 m/s² acceleration and printer height only about 0.6 m.

The maximum moment of the stepper motor is also quite a limit when trying to achieve low dynamic error (low ringing). E.g. using a Nema23 stepper (103 Ncm) would lead to maximum acceleration of 12.8m m/s² while still having the errors in the range of 14 µm.
If you have ringing and do not drive your steppers with the maximum current the steppers and the stepper drivers allow then increase the driving current. It may help if some other errors are not much larger.

Just to give you an idea where the ringing comes from. Diagonal rods have minimum influence on it (if any at all).

Edit: Edits in italic.

Edited 1 time(s). Last edit at 02/23/2018 02:43AM by hercek.

Quote
hercek
The maximum moment of the stepper motor is also quite a limit when trying to achieve low dynamic error (low ringing). E.g. using a Nema23 stepper (103 Ncm) would lead to maximum acceleration of 12.8m m/s² while still having the errors in the range of 14 µm.
If you have ringing and do not drive your steppers with the maximum current the steppers and the stepper drivers allow then increase the driving current. It may help if some other errors are not much larger.

0.9deg motors are better than 1.8deg motors here, because for the same holding torque they are twice as stiff as 1.8deg motors. In practice 0.9deg motors usually have a little less holding torque than similar 1.8deg motors, so the gain isn't quite double.

---

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
0.9deg motors are better than 1.8deg motors here, because for the same holding torque they are twice as stiff as 1.8deg motors. In practice 0.9deg motors usually have a little less holding torque than similar 1.8deg motors, so the gain isn't quite double.

I agree. Both Nema17 and Nema23 used in the analyzes were 0.9°. Micro-stepping decreases the holding torque very badly so it is better to use 0.9° motor and coarser microstepping than 1.8° stepper with finer microstepping (if their full step holding torque is about the same). From what I have seen, 1.8° motors have slightly higher holding torque than 0.9° motors but when lower microstepping of 1.8° motor is taken into account then the situation changes and 0.9° motor turns out better.

Edited 1 time(s). Last edit at 02/22/2018 06:26AM by hercek.

50% of the belt is unused, so why don't we replace it with something less stretchy, like spectra line or steel wire?

Quote
DjDemonD
Tell us more about the belts? Are they stretching?

At 120mm/s they do stretch, but I didn't see much ghosting. More like overshooting the corners.

I just swapped on some genuine gates 2gt power grip belts can't say it makes any difference still see a little ghosting when printing reasonably fast. Had no name ebay belts before.

Edited 1 time(s). Last edit at 02/22/2018 08:47AM by DjDemonD.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

Young's modulus of spectra is about 120 GPa. Steel is 200 GPa. Glass is about 70 GPa. Spectra cable needs to be 1.3 times thicker to stretch the same as a steel cable. My steel core belts have steel cross-section area of 0.22 mm². Your Spectra cable would need cross-section area at least 0.37 mm² (diameter 0.7 mm) to be comparable to my steel core belt.

Anyway you can almost halve the belt stretch error by replacing half of it with something much stiffer than the belt. Sounds like an interesting idea.

For people who have classical rostock with Ø 8 mm smooth rods and care about dynamic errors:
Ø 8 mm smooth rods bend too easily. Replace with at least Ø 12 mm smooth rods.

Quote
hercek
Micro-stepping decreases the holding torque very badly...

Only if you define holding torque as the torque needed to move the motor one microstep from the position it is trying to maintain. But holding torque in not normally defined that way. It is normally defined as either the torque needed to move the motor by one full step, or the torque needed to make the motor skip a step. Under both definitions, holding torque is independent of microstepping after you have gone from full to half stepping.

---

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].

10mm hollow, the majority of the stiffness of a cylinderical shaft comes from the "outer" diameter.

Even if you increase the stiffness of the belt , what happens is that the ringing frequency increases and the amplitude decreases, but the duration of the "ringing" might be longer so it might be a zero sum game.
And if you think about it even if you have infinite belt stiffness, what is stopping the stepper motor from ringing? the only thing preventing the stepper motor from turning is the electromagnetic force, why would you assume the stepper motor stiffness is more than the belt stiffness.

dc42: Ok, I defined holding torque as torque needed to hold the given microstep position.

Vigilant: That's why I said one needs also high holding torque and recommended to increase stepper motor current to the maximum allowed. If the maximum acceleration set in firmware does not need more (micro-step adjusted) holding torque than the motor can provide then the motor will ring at most one micro-step.

Edited 1 time(s). Last edit at 02/23/2018 02:43AM by hercek.

Well, acceleration definitely makes all the difference. I printed two toy rockets.

One with acceleration 3000 on X and Y:



And one with it set to 1500:


(there isn't a layer shift here the part detached and I printed the top 16mm separately and bonded it together)

There is much less ghosting on the second.

Now I do also have a screw driven delta but its a bit too new and untuned, to use it to test to see if screws eliminate this effect, and its belts.

Should tightening the belts really tight eliminate this?

Edited 1 time(s). Last edit at 02/23/2018 07:21AM by DjDemonD.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

Acceleration is more important than speed when considering ringing.
Ringing depends about linearly on acceleration. The more acceleration the more ringing.
Speed only needs to be above some minimum for ringing to appear. After that minimum the ringing does not really depend on speed.

Your lead screw delta probably will not be able to achieve high enough accelerations (a probably not even speed) to produce some ringing. The screws will be stiffer than belts but you will get play at the screw-nut connection and you still get some elasticity at steppers and elasticity/play at towers.

If you mind ringing you need stiffer belts, stronger motors and probably also one of:

• stiffer smooth rods (if you used them on towers)
• stiffer rollers (if you use rollers on carriages).

Linear rails will be stiff enough but I never found specifications how precise they are (how much play do they have). This value is typically specified for smooth rods and linear bearings but it is not specified for linear rails.

I have linear rails on the XL, gates belts 6mm 2gt, Haydn rods 360mm 6mm diameter, smart effector and carriages. Its a 2040 frame with robot digg corners. So it has to be the belts, or the carbon rods.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

It surprises me, that reducing acceleration alone changes ringing? To me, the jerk settings are more important, since the ringing happens right in "jerk-zone". Acceleration kicks in later, I thought?
I'll test my ringing testpart with reduced acceleration and then with lower jerk, see what happens.

Edited 1 time(s). Last edit at 02/24/2018 04:16AM by o_lampe.

Acceleration and jerk can both contribute to ringing. Jerk is essentially a period of high acceleration that lasts a very short time; so unless you have your jerk set very high, reducing it will probably have less effect than reducing acceleration. Also, all transitions between consecutive moves in different directions use acceleration and deceleration, but only some use jerk.

---

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
DjDemonD
I have linear rails on the XL, gates belts 6mm 2gt, Haydn rods 360mm 6mm diameter, smart effector and carriages. Its a 2040 frame with robot digg corners. So it has to be the belts, or the carbon rods.

If your Ø 6 mm diagonal (carbon) rods do not have any noticeable play at the joints and the joints move easily then your diagonal rods are not a problem.

Unless they bend under acceleration. A 6mm carbon rod 360mm long is stiff but you can bend it.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

Of course you can bent it. But the important thing is that a diagonal rod is loaded only in the direction of the axe along its length (if your ball joints are free to move). Try to bend your rods by pressing them from their ends. Use force only in the direction of the axe along the rod length. Notice that the forces exerted by the printer on your diagonal rods are in the range of 7 N.

The only way how diagonal rods could be an issue is if the ball joints do not move freely or when the ball joints are not centered on the axe along the rod length.

Some numbers so that you have a rough idea what we are talking about:
For a carbon fiber laminate rod long 36 cm, 6 mm external diameter, 1 mm wall thickness:
Compressive strength about 1.6 GPa, Young's modulus about 70 GPa (depends on material and grain direction, some go up to 180 GPa).
Error due to compression / elongation of the rod under force of 7 N is 2 µm (totally negligible compared to belts).
Compression pressure is 446 kPa (about 3587 times smaller than the compression strength).
The rod can be considered medium or even a long beam so we must consider buckling. Buckling would start at 544 N (about 78 times higher than our force of 7 N).

Although these are theoretical numbers for an ideal carbon rod with an ideal load ... still they have such a huge safety margins that rod is unlikely a problem when dealing with ringing in your prints.

That sir is a solid explanation. Thank you. I guess we rule out the rods then.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

This is such a wonderful discussion -- I love solid engineering!

On my printers I've been seeking to reduce ringing.
The first thing I did was add diagonal bracing to the frame, which made it a lot stiffer.

Second, I spoke to the engineer at Goodwinds (who produce the CF tubing) and he said that for this application, the 0.230" strong walled CF tubing is more than stiff enough.
People sometimes choose 10mm CF tubing, but it is overkill and mostly for aesthetic reasons.
Note: 10mm CF tubing is quite a bit more expensive, and would dominate the cost of the arms. Why spend substantial money where it won't do noticeable good?

Then I realized that my carriages and belt tighteners were slightly stretchy.
I stopped using zip ties for belt tensioning, and switched to using tensioners which rely on two M3x35 bolts. They're precise and not stretchy. This alone made the inside diameters of some of the holes in my parts increase by 0.05mm, because they got smoother.
To improve the carriages, I switched to using larger W-wheels riding on the outside corners, to give them a wider basis to resist twisting, and I modified the carriages so that they have both an inside and an outside piece, so the bearing axles are not longer cantilevered.

On my latest printer I'm building it with 9mm belts and using 4 wheel carriages, to see if that has any further benefits.

Here is a link my Kumu-3D printer: [github.com]
I'm still working on the assembly manual: [docs.google.com]

I've done a bit of testing, mostly printing 3DBenchy's using Smoothie, and for the ringing I observed, stiffening the printer helped, and reducing the junction deviation and acceleration also helped. If I remember correctly, reducing the junction deviation was more important than the acceleration, but that may be particular to Smoothie. Soon I hope to do some testing with RRF on a Duet. :-)

Quote
DjDemonD
But I do need to make some new rods for the fishtank printer in abs or petg as it gets quite hot inside, if I can sort the leadscrews out first. So I could try the 10mm carbon tubes (assuming they fit over the Haydn rod ends?) and then I could swap rods and do some fairly well controlled testing.

Hi DjDemonD,

If your printing environment might get above 80°C, then you're going to have other problems. The super-strong N52 magnets I use shouldn't be heated to above 80°C, because otherwise they will permanently loose some strength.
It is possible to get N50M magnets which can go up to 100°C: [www.kjmagnetics.com]
If you want to go warmer, N48H will go to 120°C.
Magnetic strength is directly proportional to the rating number, so N50H is about 96% of the strength of N52, etc.

Note: I recently asked the engineer at Goodwinds what the maximum temperature their CF tubing can stand, and they said 220°F, which is about 104.4°C. But since you're in a metric country, you could always shift to using 10mm OD aluminum tubing.

For this application, one benefit of using 10mm OD tubing is that it is the same diameter as the magnets, so you can completely eliminate the need for using PLA, ABS, or PETG adapters between the tubing and the Delrin rod ends.

My guess is that if you use leadscrews in a delta printer, you'll find that even though the leadscrews themselves are stiff, the anti-backlash nuts use springs, which aren't terribly stiff. Also, usually leadscrews have more mass, and thus limit the acceleration/speed a bit.

I'm only aiming for around 45 deg C, the fishtank print gets up to about 50 deg C at ABS temps, it is very much enclosed, not loosely like most enclosures (not quite airtight but not far off). That's hot enough for me and for warp free abs parts. My home made PLA rods have been working well. Although the machine only calibrates to 0.2mm for an adapted fishtank I can live with that. I;ve got an XL for really precise stuff.

I'm using N50 magnets and they seem to be up to the job. If the machine were more serious (like the XL) I;d buy some rods from you they really are made extremely well, but its more of an experiment into the unusual and won't ever be a workhorse machine.

The 10mm tubing would look cool on that machine and the arms are only 220mm so not too expensive, although I can only find pulltruded not roll wrapped so visually they won't look so good.

I haven't fitted antibacklash nuts yet thats a job I have to do over the next few days I've got some now and have to print some redesigned carriages for them.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions