Printed Arms: print them before the incident!
Posted by PRZ
|
Printed Arms: print them before the incident! September 03, 2015 05:53PM |
Yesterday, while testing some M codes, I heard a bad noise and duh!, one arm was broken.
However, I was quite lucky, as I printed the day before some prototypes for printed arms, for future tests.
So, I had no other option than install and test these prototype arms.
Mixed with the original arms, they worked for printing, but I had poor calibration.
Looking closely, you could see that the printed ends are larger at the fork root than the printed end in operation, for tighter spring effect (it nearly doubled the load on the ball)
Anyway, mixing steel and plastic arms is not very good because the thermal expansion is quite different and with my 'light spot' heated bed, I will have temperature size change, so I decided to change all arms, and also to make them longer, for better large diameter printing (which does not work well on Fisher with original arms). I lengthen them from 160 to 170mm. I lost 11mm of height, but the maximum diameter is now workable, with no step loss. However, as arms are thinner, I did not loose maximum diameter height (150mm diameter maximum height is around 83mm, this is limited by the clearance between arms and panels)
In fact, I discover afterward that my poor calibration was due to a hole in the Buildtak at the center calibration point, so my conclusion that this was related to the new arms was quite premature.
Have a look on the BuildTak, which really needs to be retired! The small hole in the middle was the culprit for the faulty calibration, not the new arms.
How are they designed:
They are two printed ends fork-shaped joined by a M3 threaded rod. The length adjustment is done by firmly pressing the internal nut against a thin ring. This allow precise length adjustment (~0,01mm).
As you could see, there is a hole to set a screw pressing the flanges against the ball, but this is not needed, the pinch of the last version being really firm.
They are printed in PETG and I doubt that PLA could be usable, it may broke while setting on balls. ABS is most probably workable.
I initially printed end with cups but they were difficult to fit and the main problem is that it is quite difficult to measure axis to axis length and so, to adjust the arm length. Also, that needs springs.
Compared to the original arms, the holes are shaped to the ball size and the contact area is then much larger. The overall stiffness looks comparable to the original arms, and with calibration spot moved away from holes, I am back to less than 0.01mm difference on 7 points calibration, as before.
The arms are slightly lighter than the original (20g the pair instead of 25 g). the 170mm arm pair weight 22g.
If you decide to lengthen the arms, do not forget to modify and reload the config.g file, then shut down and restart the machine. Do not home before restart!
For 170mm arms, the homed height shall be modified to 171mm.
Threaded M3 rod length is 118mm for 160mm arms, 128mm for 170mm arms. 14mm insertion in simple end, 19mm insertion in nut end. You shall clean well the threaded rod end. Beware not to bend the rods while cutting them if using a manual saw.
Medium washer is required for proper compression adjustment.
Clean the angel hair (I have a lot of angel hair because this is printed with esun PETG which need very high temperature. Lower temperature PETG may be better).
I printed with 0.25mm layers, but you may prefer to print with 0.2mm layers with a standard Fisher (mine have a geared extruder). Slow down speed to 60mm/s. Fill in 70%, honeycomb. This is a quite long print, total may be two hours. One arm set print time is less than half an hour.
You may print a first arm set to check, then the 5 others.
I recommend everyone to have at least one arm set in case of unexpected mechanical incident. (I was quite surprised by the event, and that break instantly!)
I printed a calibration angle of 100mm, which is near the maximum size (that makes 145 mm on the diagonal with the peripheral loop), and that worked perfectly, without the steps loss that I experienced on larger parts before (I used a travel speed reduced to 180 mm/s).
To fit the ends on balls, push them on the screw, then slide to the ball, that is easier than direct fit.
It is better to set the end with the nut on the carriage for weight reasons, but this is not critical.
[edit] Don't forget to lubricate:
- If printed in ABS, use plastic compatible lubricant. [forums.reprap.org]
- If printed in PETG, you could use any lubricant, as PETG is chemically quite resistant
File with STL and OpenScad in next message :
Arm length measurement will be explained in next message.
Edited 3 time(s). Last edit at 09/03/2015 07:03PM by PRZ.
However, I was quite lucky, as I printed the day before some prototypes for printed arms, for future tests.
So, I had no other option than install and test these prototype arms.
Mixed with the original arms, they worked for printing, but I had poor calibration.
Looking closely, you could see that the printed ends are larger at the fork root than the printed end in operation, for tighter spring effect (it nearly doubled the load on the ball)
Anyway, mixing steel and plastic arms is not very good because the thermal expansion is quite different and with my 'light spot' heated bed, I will have temperature size change, so I decided to change all arms, and also to make them longer, for better large diameter printing (which does not work well on Fisher with original arms). I lengthen them from 160 to 170mm. I lost 11mm of height, but the maximum diameter is now workable, with no step loss. However, as arms are thinner, I did not loose maximum diameter height (150mm diameter maximum height is around 83mm, this is limited by the clearance between arms and panels)
In fact, I discover afterward that my poor calibration was due to a hole in the Buildtak at the center calibration point, so my conclusion that this was related to the new arms was quite premature.
Have a look on the BuildTak, which really needs to be retired! The small hole in the middle was the culprit for the faulty calibration, not the new arms.
How are they designed:
They are two printed ends fork-shaped joined by a M3 threaded rod. The length adjustment is done by firmly pressing the internal nut against a thin ring. This allow precise length adjustment (~0,01mm).
As you could see, there is a hole to set a screw pressing the flanges against the ball, but this is not needed, the pinch of the last version being really firm.
They are printed in PETG and I doubt that PLA could be usable, it may broke while setting on balls. ABS is most probably workable.
I initially printed end with cups but they were difficult to fit and the main problem is that it is quite difficult to measure axis to axis length and so, to adjust the arm length. Also, that needs springs.
Compared to the original arms, the holes are shaped to the ball size and the contact area is then much larger. The overall stiffness looks comparable to the original arms, and with calibration spot moved away from holes, I am back to less than 0.01mm difference on 7 points calibration, as before.
The arms are slightly lighter than the original (20g the pair instead of 25 g). the 170mm arm pair weight 22g.
If you decide to lengthen the arms, do not forget to modify and reload the config.g file, then shut down and restart the machine. Do not home before restart!
For 170mm arms, the homed height shall be modified to 171mm.
Threaded M3 rod length is 118mm for 160mm arms, 128mm for 170mm arms. 14mm insertion in simple end, 19mm insertion in nut end. You shall clean well the threaded rod end. Beware not to bend the rods while cutting them if using a manual saw.
Medium washer is required for proper compression adjustment.
Clean the angel hair (I have a lot of angel hair because this is printed with esun PETG which need very high temperature. Lower temperature PETG may be better).
I printed with 0.25mm layers, but you may prefer to print with 0.2mm layers with a standard Fisher (mine have a geared extruder). Slow down speed to 60mm/s. Fill in 70%, honeycomb. This is a quite long print, total may be two hours. One arm set print time is less than half an hour.
You may print a first arm set to check, then the 5 others.
I recommend everyone to have at least one arm set in case of unexpected mechanical incident. (I was quite surprised by the event, and that break instantly!)
I printed a calibration angle of 100mm, which is near the maximum size (that makes 145 mm on the diagonal with the peripheral loop), and that worked perfectly, without the steps loss that I experienced on larger parts before (I used a travel speed reduced to 180 mm/s).
To fit the ends on balls, push them on the screw, then slide to the ball, that is easier than direct fit.
It is better to set the end with the nut on the carriage for weight reasons, but this is not critical.
[edit] Don't forget to lubricate:
- If printed in ABS, use plastic compatible lubricant. [forums.reprap.org]
- If printed in PETG, you could use any lubricant, as PETG is chemically quite resistant
File with STL and OpenScad in next message :
Arm length measurement will be explained in next message.
Edited 3 time(s). Last edit at 09/03/2015 07:03PM by PRZ.
|
Re: Printed Arms: print them before the incident! September 03, 2015 06:08PM |
Here, the STL and OpenScad files
OpenScad and STL for 1 set
Fisher_arms1.zip
STL for 5 sets
Fisher_arms5.zip
To measured accurately the arm length, you shall use 4mm flush rods, which fits in the end holes
M3 thread pitch is 0.5 mm, so a half turn will move the end by 0.25mm. The ensemble shall be tighten firmly, so the initial length shall be shorter of more than 0.5mm than the desired length. In fact, at initial setup, the nut is not pushed against the ring, so you may need to set the initial length even shorter than that.
The rod is screwed in the simple end. Make a mark to not overtight.
The rod shall not screw in the end with the nut, as this prevent the ring compression to work properly. Clean the hole if needed.
Always measure with another arm set at the end of flush rods for proper alignment. I feel more simple and accurate to measure the inside of the rods than the outside.
If you only have 150mm caliper, use a calibrated part to compensate the width difference (see photo).
[edit] You shall take into account the rod diameter (4mm), and if needed, the caliper internal measure space - say, with the shown vernier caliper, the measure was : 170-4-10 = 156mm. If you have caliper with crossed internal measure jaws, as the digital caliper shown, the internal measure space is 0.
The arm length is not very critical (you will set it in the config.g) , but all your arms shall have the SAME length, with the best precision you can obtain, and in any case, better than 0.02mm ([edit] that is better done with the vernier caliper that I have used to build my set than with a digital caliper).
[edit] Remember that if the resolution of digital calipers is 0.01mm, their accuracy is only 0.02mm, however their repeatability is probably better than their accuracy.
Edited 2 time(s). Last edit at 09/03/2015 07:25PM by PRZ.
OpenScad and STL for 1 set
Fisher_arms1.zip
STL for 5 sets
Fisher_arms5.zip
To measured accurately the arm length, you shall use 4mm flush rods, which fits in the end holes
M3 thread pitch is 0.5 mm, so a half turn will move the end by 0.25mm. The ensemble shall be tighten firmly, so the initial length shall be shorter of more than 0.5mm than the desired length. In fact, at initial setup, the nut is not pushed against the ring, so you may need to set the initial length even shorter than that.
The rod is screwed in the simple end. Make a mark to not overtight.
The rod shall not screw in the end with the nut, as this prevent the ring compression to work properly. Clean the hole if needed.
Always measure with another arm set at the end of flush rods for proper alignment. I feel more simple and accurate to measure the inside of the rods than the outside.
If you only have 150mm caliper, use a calibrated part to compensate the width difference (see photo).
[edit] You shall take into account the rod diameter (4mm), and if needed, the caliper internal measure space - say, with the shown vernier caliper, the measure was : 170-4-10 = 156mm. If you have caliper with crossed internal measure jaws, as the digital caliper shown, the internal measure space is 0.
The arm length is not very critical (you will set it in the config.g) , but all your arms shall have the SAME length, with the best precision you can obtain, and in any case, better than 0.02mm ([edit] that is better done with the vernier caliper that I have used to build my set than with a digital caliper).
[edit] Remember that if the resolution of digital calipers is 0.01mm, their accuracy is only 0.02mm, however their repeatability is probably better than their accuracy.
Edited 2 time(s). Last edit at 09/03/2015 07:25PM by PRZ.
|
Re: Printed Arms: print them before the incident! September 04, 2015 07:10AM |
7 points calibration, direct from start. This is with my ravaged Buildtak !
below 0.01 is normal, that was the same before changing the arms. I even get sometimes 0 (I got it one time with the new arms)
You see also the new parameters.
With properly defined switch offset and other parameters, I am normally below 0.2 before calibration. Here that was a bit exceptional.
What is the trick ?
- properly tensioned belts (not too much, however, that increase friction)
- lower bed spring pressure - that is due to the change to aluminium bed
That gives a quite low offset difference between center and near column probe points. I set 0.05.
But the trick is that you shall take into account the intermediate offset for the half-way probe points.
so I set 0.03 difference for the halfway offset and 0.05 for the centre probe offset.
Note that the probe points are slightly moved to take into account the poor state of the buildtak.
[edit] I also reduced the probe test clearance to 2mm instead of 4, which reduce calibration time.
For good arm length measurement, my 40 years old vernier caliper rules !
Edited 1 time(s). Last edit at 09/04/2015 07:12AM by PRZ.
below 0.01 is normal, that was the same before changing the arms. I even get sometimes 0 (I got it one time with the new arms)
You see also the new parameters.
With properly defined switch offset and other parameters, I am normally below 0.2 before calibration. Here that was a bit exceptional.
What is the trick ?
- properly tensioned belts (not too much, however, that increase friction)
- lower bed spring pressure - that is due to the change to aluminium bed
That gives a quite low offset difference between center and near column probe points. I set 0.05.
But the trick is that you shall take into account the intermediate offset for the half-way probe points.
so I set 0.03 difference for the halfway offset and 0.05 for the centre probe offset.
Note that the probe points are slightly moved to take into account the poor state of the buildtak.
[edit] I also reduced the probe test clearance to 2mm instead of 4, which reduce calibration time.
For good arm length measurement, my 40 years old vernier caliper rules !
Edited 1 time(s). Last edit at 09/04/2015 07:12AM by PRZ.
|
Re: Printed Arms: print them before the incident! September 04, 2015 08:29AM |
[EDIT : wrong calibration routine, see next post]
I have taken the 3 other probe points from your post, and the results are less brilliant, while they are relatively constant.
However, while probing 10 points, the message still say this is a 7 points calibration, so I don't know if it takes the seven last probe or if this is an erroneous message ?
Best result is 0.031, worse 0.037.
Iterating the probe macro always worsen the results, so all the tests are done after restart.
This probing gives less angles on the columns, which I think is more accurate.
Here is the probing sequence. It shall be noted that I obtained the best results with an offset sligthly higher for the inter-column points than for the center point (by 0.01).
Edited 1 time(s). Last edit at 09/04/2015 09:32AM by PRZ.
I have taken the 3 other probe points from your post, and the results are less brilliant, while they are relatively constant.
However, while probing 10 points, the message still say this is a 7 points calibration, so I don't know if it takes the seven last probe or if this is an erroneous message ?
Best result is 0.031, worse 0.037.
Iterating the probe macro always worsen the results, so all the tests are done after restart.
This probing gives less angles on the columns, which I think is more accurate.
Here is the probing sequence. It shall be noted that I obtained the best results with an offset sligthly higher for the inter-column points than for the center point (by 0.01).
Edited 1 time(s). Last edit at 09/04/2015 09:32AM by PRZ.
|
Re: Printed Arms: print them before the incident! September 04, 2015 09:31AM |
Ok, I found my mistake in the 10 points calibration routine, my probe points were not properly numbered.
The 10 points calibration deviation is remarquably constant at 0.043 (4 different tests).
The corrected calibration routine (Pn numbers , they shall go from P0 to P9)
I slightly increased the offset for inter-columns points at 0.22 (for 0.2 at center), which gives better results
Edited 2 time(s). Last edit at 09/04/2015 10:50AM by PRZ.
The 10 points calibration deviation is remarquably constant at 0.043 (4 different tests).
The corrected calibration routine (Pn numbers , they shall go from P0 to P9)
I slightly increased the offset for inter-columns points at 0.22 (for 0.2 at center), which gives better results
Edited 2 time(s). Last edit at 09/04/2015 10:50AM by PRZ.
|
Re: Printed Arms: print them before the incident! September 05, 2015 06:29AM |
Nice, that's the modification I've done yesterday 
Thanks
Thanks
Sorry, only registered users may post in this forum.