Y axis stutter on MendelMax

I have a mendelmax with the wooden cut Y bed and I'm encountering a bit of shaking on the y axis. When it moves sometimes it doesn't look smooth and although it seems to print ok I can see a little bit of waviness caused by the shake of the y axis. I tried adjusting the current but nothing seems to fix it and the bed seems to be reasonably smooth.

I have:

48nM steppers
Polulu 4988 w/RAMPS 1.4
Arduino Mega 2560
GT2 belts
8mm self centering bushings



Edited 1 time(s). Last edit at 05/07/2012 12:15PM by LT1Pat.

did you disconnect the belt and see how smooth it was when you moved it by hand?

It moves really smooth by hand which is the strangest part. I can glide the Y axis along by giving it a push with the belt removed. I played with the acceleration numbers and switches axis and the problem persists. Could the stepper still function if it were wired incorrectly?

When a part is printed the ripples from the jerky Y axis show up on any parts that are parallel to the axis. The X axis is very smooth and as you can see in the picture the jerks are pretty consistent layer to layer.

How does it move with the belt on and the stepper un powered?
There will obviously be resistance, but do you feel anything other than the stepper steps?
Assuming it feels as you would expect, then it's probably not a mechanical issue, so the next thing I'd try is swapping the X/Y wiring at the control board, if that didn't work I'd try swapping the steppers as well.

Are you using the laser-cut Y carriage with the aluminum bed? I had this same issue. The carriage is overconstrained; if the pieces aren't perfect (mine weren't) it creates a slight twist which causes binding. The binding from the twist can't be fixed using the brass bushings' adjustment feature. Also, the belt attaches at a point which causes the carriage to flex the brass bushings into a different angle, causing them to bind even more.

Here's what I did:
* I removed two cross pieces (I think they were numbered "4"; I'm not sure). This loosened it up enough where I could forcefully twist it into alignment. I had to adjust the carriage, the bushings (also takes force), the tightness of the 4 screws to the bed, and the positions of the rods to the frame. This takes a lot of trial, error, and careful measurement with a dial indicator.

* I drastically reduced acceleration to 1000 mm/s^2 (my Prusa was at 3000). Higher acceleration causes binding from flexing.

Todd

I tried swapping drives so the Y axis became the X axis and the X axis became the Y and the problem stayed with the Y axis. The bed moves pretty smooth but it seems to change with how much the bed is heated, I think it starts to constrain when it's hot as if all the wood is shifting.

Polygonhell Wrote:
-------------------------------------------------------
> How does it move with the belt on and the stepper
> un powered?
> There will obviously be resistance, but do you
> feel anything other than the stepper steps?
> Assuming it feels as you would expect, then it's
> probably not a mechanical issue, so the next thing
> I'd try is swapping the X/Y wiring at the control
> board, if that didn't work I'd try swapping the
> steppers as well.

Yes I am using the laser cut Y carriage and the new super flat aluminum bed. I think you're really onto something because I can stare at the stepper pulley intensely and I can't seem to see any sort of jerking or vibration that I'm actually seeing in the bed. I would say that it seems more like a bed chatter then an actual electronic or stepper issue. I took the bed off multiple times and re-aligned the bushings to the point where it just glides across with the flick of the finger but as the bed heats up I wonder if it changes because I can't really tell with the belt attached since I'm feeling the stepper motor resistance. The chatter (although silent) seems to change every time I adjust the bed or feed rates so I can't really find a real strong correlation. I took a few pictures of some different chatter and some are much worse than others but all are seriously killing any sort of quality.

The first on is the 40mm calibration cube with some decent feed rates, the second one is the same cube with slower feed rates, the third one is a cube with some curves with the same feed rates as the first one.





tbfleming Wrote:
-------------------------------------------------------
> Are you using the laser-cut Y carriage with the
> aluminum bed? I had this same issue. The carriage
> is overconstrained; if the pieces aren't perfect
> (mine weren't) it creates a slight twist which
> causes binding. The binding from the twist can't
> be fixed using the brass bushings' adjustment
> feature. Also, the belt attaches at a point which
> causes the carriage to flex the brass bushings
> into a different angle, causing them to bind even
> more.
>
> Here's what I did:
> * I removed two cross pieces (I think they were
> numbered "4"; I'm not sure). This loosened it up
> enough where I could forcefully twist it into
> alignment. I had to adjust the carriage, the
> bushings (also takes force), the tightness of the
> 4 screws to the bed, and the positions of the rods
> to the frame. This takes a lot of trial, error,
> and careful measurement with a dial indicator.
>
> * I drastically reduced acceleration to 1000
> mm/s^2 (my Prusa was at 3000). Higher acceleration
> causes binding from flexing.
>
> Todd

It sounds like you're past the hardest part (alignment). Here's the technique I used to tune acceleration (I assume you're using Marlin). I created a simple g-code file:

G1 Y100 F5400
G1 Y100.7 F5400
(repeat these two lines hundreds of times)

Your numbers will probably be different. I started with different numbers then adjusted them until I hit the worst case: the motor and bed stood still (this is a form of resonance). The goal is to reduce acceleration to the point where it can't resonate.

While it's resonating, paste commands into Pronterface to adjust the acceleration (Pronterface interjects manual commands into the running gcode )

M204 S1000.00

You want the highest value that allows it to move back and forth the full distance without overshooting. Once you've found it, use the following to permanently store the adjustment:

M500

If you still get stuttering on real gcode files then you might need to reduce it farther. When you're using Marlin and a heavy bed, the most important number isn't the top speed, it's the acceleration. Acceleration controls the forces on the machine.

Todd

This sounds like a great idea, I'm going to try this out tomorrow. Are you adjusting the feed (5400 number) and the acceleration or just the acceleration? I'm also going to strip the rods of all oil as well as the bearings because mendelmax says not to use any but I originally didn't use any oil and it was still really bad. Are you saying that M500 would permanently store the adjustment in the Arduino board or just until you close pronterface?

Thanks!

tbfleming Wrote:
-------------------------------------------------------
> It sounds like you're past the hardest part
> (alignment). Here's the technique I used to tune
> acceleration (I assume you're using Marlin). I
> created a simple g-code file:
>
> G1 Y100 F5400
> G1 Y100.7 F5400
> (repeat these two lines hundreds of times)
>
> Your numbers will probably be different. I started
> with different numbers then adjusted them until I
> hit the worst case: the motor and bed stood still
> (this is a form of resonance). The goal is to
> reduce acceleration to the point where it can't
> resonate.
>
> While it's resonating, paste commands into
> Pronterface to adjust the acceleration
> (Pronterface interjects manual commands into the
> running gcode )
>
> M204 S1000.00
>
> You want the highest value that allows it to move
> back and forth the full distance without
> overshooting. Once you've found it, use the
> following to permanently store the adjustment:
>
> M500
>
> If you still get stuttering on real gcode files
> then you might need to reduce it farther. When
> you're using Marlin and a heavy bed, the most
> important number isn't the top speed, it's the
> acceleration. Acceleration controls the forces on
> the machine.
>
> Todd

I'm pretty sure I understand what's actually happening.

1.) The belt pulls on the Y-carriage and tweaks it momentarily due the high inertia of the heavy aluminum bed
2.) The bed binds for a fraction of a millisecond and the belt stretches
3.) The bed releases and the increased tension of the belt causes a slight jump
4.) Tightening the belts help the problem I but tweaks the bed more which just changes the frequency of the vibrations/stutter

Take this and rinse/repeat very rapidly. I did notice that the GT2 belts that come with the mendelmax do stretch which may be part of the problem that allows this harmonic to take place. Adjusting the belt tension and bed bushing alignment as well as feedrates seems to change the frequency of the harmonic enough to make you scratch your head.

Possible solutions:

1.) Gusset for the wooden bed to stop flexing from where the belt attached to somewhere
2.) Belt's that don't stretch as much or wider
3.) Lighter bed (but I love the bed!)
4.) Re-align bushings while belts are under a decent amount of tension
5.) Unconstrain the bed somehow

LT1Pat Wrote:
-------------------------------------------------------
> This sounds like a great idea, I'm going to try
> this out tomorrow. Are you adjusting the feed
> (5400 number) and the acceleration or just the
> acceleration? I'm also going to strip the rods of
> all oil as well as the bearings because mendelmax
> says not to use any but I originally didn't use
> any oil and it was still really bad. Are you
> saying that M500 would permanently store the
> adjustment in the Arduino board or just until you
> close pronterface?
>
> Thanks!

I adjust the feed (5400) and the distance (.7) until it resonates (stops moving). I then adjust the acceleration (1000) to prevent the resonance. M500 stores it into the Arduino.

LT1Pat Wrote:
-------------------------------------------------------
> I'm pretty sure I understand what's actually
> happening.
>
> 1.) The belt pulls on the Y-carriage and tweaks it
> momentarily due the high inertia of the heavy
> aluminum bed
> 2.) The bed binds for a fraction of a millisecond
> and the belt stretches
> 3.) The bed releases and the increased tension of
> the belt causes a slight jump
> 4.) Tightening the belts help the problem I but
> tweaks the bed more which just changes the
> frequency of the vibrations/stutter

This matches the conclusion I reached while working on mine.

>
> Take this and rinse/repeat very rapidly. I did
> notice that the GT2 belts that come with the
> mendelmax do stretch which may be part of the
> problem that allows this harmonic to take place.
> Adjusting the belt tension and bed bushing
> alignment as well as feedrates seems to change the
> frequency of the harmonic enough to make you
> scratch your head.
>
> Possible solutions:
>
> 1.) Gusset for the wooden bed to stop flexing from
> where the belt attached to somewhere
> 2.) Belt's that don't stretch as much or wider
> 3.) Lighter bed (but I love the bed!)
> 4.) Re-align bushings while belts are under a
> decent amount of tension
> 5.) Unconstrain the bed somehow

These may help, but I'd wait until getting the accelleration adjusted. After that, I'd be curious how much higher the accelleration can be after these changes.

Todd

I just thought of another solution, I could tension the area between the two points with extra belt where the belt attach in order to neutralize the static load that is tweaking the bed. I think I am also going to implement this tomorrow...

I tried this and no matter what I do it seems to be able to handle it. I think I have the max acceleration in the firmware limited though so that might be stopping it. I preloaded the area between the two belt attachment points which may have fixed it. I'm printing a test cube right now to take a look.


LT1Pat Wrote:
-------------------------------------------------------
> This sounds like a great idea, I'm going to try
> this out tomorrow. Are you adjusting the feed
> (5400 number) and the acceleration or just the
> acceleration? I'm also going to strip the rods of
> all oil as well as the bearings because mendelmax
> says not to use any but I originally didn't use
> any oil and it was still really bad. Are you
> saying that M500 would permanently store the
> adjustment in the Arduino board or just until you
> close pronterface?
>
> Thanks!
>
> tbfleming Wrote:
> --------------------------------------------------
> -----
> > It sounds like you're past the hardest part
> > (alignment). Here's the technique I used to
> tune
> > acceleration (I assume you're using Marlin). I
> > created a simple g-code file:
> >
> > G1 Y100 F5400
> > G1 Y100.7 F5400
> > (repeat these two lines hundreds of times)
> >
> > Your numbers will probably be different. I
> started
> > with different numbers then adjusted them until
> I
> > hit the worst case: the motor and bed stood
> still
> > (this is a form of resonance). The goal is to
> > reduce acceleration to the point where it can't
> > resonate.
> >
> > While it's resonating, paste commands into
> > Pronterface to adjust the acceleration
> > (Pronterface interjects manual commands into
> the
> > running gcode )
> >
> > M204 S1000.00
> >
> > You want the highest value that allows it to
> move
> > back and forth the full distance without
> > overshooting. Once you've found it, use the
> > following to permanently store the adjustment:
> >
> > M500
> >
> > If you still get stuttering on real gcode files
> > then you might need to reduce it farther. When
> > you're using Marlin and a heavy bed, the most
> > important number isn't the top speed, it's the
> > acceleration. Acceleration controls the forces
> on
> > the machine.
> >
> > Todd

I tried loading the code but it was able to handle everything up to the max feedrate numbers and any acceleration I put in it. I tried tightening the belts very tight and preloading the bed and it seemed to help but it's still problematic. Here is a copy of todays print with these changes. The good pic of the side facing the front of the machine and the wavy side is the one that runs parallel with the Y axis. The stutter creates a ghosting effect.

Good:

Bad:

As the ripples are vertically aligned I don't think it is slip/go type stutter. It looks more like a pulley / belt tooth mismatch. Is the pitch of the ripples related to the belt pitch?

---

[www.hydraraptor.blogspot.com]

Check that your belts are OK. I had a similar effect when one of my belts was worn out, many of the metal wires were snapped so the belt was flexing when printing at speed and high acceleration settings I had a very similar shadow effect, disappeared when I changed the belt.

---

[richrap.blogspot.com]

This is a brand new belt

At F5400 my bed won't accelerate anything higher than S2300 without the stepper motor skipping. I counted the teeth on the pulley and they are 36 GT2 and appear to be in good shape. The belt folds over the pulley goood as far as I can tell. I just finished assembling this machine a few days ago and it has done it since the very first print.

Do you get good prints at S1000?

I did not, I set the max in the firmware to only S900 and it's just as bad as ever. The chatter if very consistent from layer to layer and it seems to change with the direction of travel. I ordered two new steppers and two new GT2 20 tooth pulleys to replace X and Y so if it's those that should be taken care of. I can look intensely at the pulley/belt as it moves and the bed shakes and I can't seem to see anything wrong with the motion of the pulley itself so it's looking like it's the bed/carriage. There is almost no resistance in the Y bed and I can glide it across the 8mm rails but they do seem to flex if I push gently on the bed.

If I went with 20MM shafting and an aluminum carriage and 4 of these: [www.vxb.com]

Would that create too much weight for the bed?

tbfleming Wrote:
-------------------------------------------------------
> Do you get good prints at S1000?

Edited 1 time(s). Last edit at 05/09/2012 08:09PM by LT1Pat.

I think I'm going to do this:

(2) of these [cgi.ebay.com]
(1) of these [cgi.ebay.com]
(1) of these [cgi.ebay.com]

Then I'm going to mill a 6061 3/8" plate and 4 spacers to join it to the mendelmax heated bed. Will this be too much weight?

I ordered all those parts so hopefully that takes care of it along with the new steppers and pulleys

Sorry to say this but I think that is way overkill. The bearings will certainly help (they are much more forgiving than bushings) but they are loud. 16mm rods are gigantic and heavy though, lm8uu in printed housings (e.g. [www.thingiverse.com]) would have sufficed. Better yet you can actually glue lm8uu into the wooden frame as the diameter is not too different.

I too suffered from the same Y axis stutter as you and there are a few discussions on the mendelmax google group as to why. I think the main points, discussed in this thread too, is the weight of the bed causing momentary binding due to a frame (and/or bushings) that is not fully square. I too went for the linear bearing route and glued them in and it did solve the problem, but I am now looking for a way to move back to bushings as they are so much quieter. I think tightening the bolts of the frame after loosening the bushings and running it with the belt might be the *right* sequence of steps, anyway I'm sure your solution will also fix the problem.

We have basically came to the same conclusion that the Y axis is over-constrained. It has a small amount of float built in to the Y shaft mount, but it is not enough to overcome the problems, and increasing the float causes enough looseness in the axis that it effects print quality. This has been tough to solve, because, apparently through sheer luck, my prototype of this Y axis worked brilliantly from day 1, with basically no binding or anything once I learned the trick of wiping the oil off the shafts. That lead me to an early assumption that the problems were simple alignment issues, and not a more fundamental design flaw. I apologize for that delay in comprehending the nature of the problem.

We are in the process of re-working the Y axis mount to eliminate the over constraint. I have the design finished in my head, but I just need a few hours to sit down in front of the computer and do up the CAD models. Things have been insane for me for the last several months, but I have allocated time this weekend to complete that design, and I will get a few sets made up and and test it next week.

Once we finish the design we will will offer a free upgrade for just the cost of postage to anyone who purchased a hardware kit or full kit from us. Barring unexpected problems with this design, I expect it to be available in about two weeks.

-Kludgineer, AKA Maxbots

Edited 1 time(s). Last edit at 05/10/2012 02:01PM by kludgineer.

I feel like part of the cause is the heavy weight of the bed hanging so far off the edge of the lightweight frame creates a moment on the entire setup which creates the vibration. Normally with reprap the 8mm is adequate because a lot of people are running some light heated PCB's and sheets of thin glass but this aluminum bed is heavy. If I lightly press down on the bed the 8mm rods are able to flex without much force at all so I think 10mm or 12 mm is required. I went with 16mm because they had those end for them that I can just tighten down with 5mm bolts instead of fabricating an end holder. Compared to the bed I don't think that the 16mm bearings will add much weight but I won't know for sure until they arrive. Hopefully the bigger bearings are quieter because I think I will miss the silence of the bushings. I also ordered some LM8UU to try what you suggested if those arrive first.

yydoctt Wrote:
-------------------------------------------------------
> Sorry to say this but I think that is way
> overkill. The bearings will certainly help (they
> are much more forgiving than bushings) but they
> are loud. 16mm rods are gigantic and heavy though,
> lm8uu in printed housings (e.g.
> [www.thingiverse.com]) would have
> sufficed. Better yet you can actually glue lm8uu
> into the wooden frame as the diameter is not too
> different.
>
> I too suffered from the same Y axis stutter as you
> and there are a few discussions on the mendelmax
> google group as to why. I think the main points,
> discussed in this thread too, is the weight of the
> bed causing momentary binding due to a frame
> (and/or bushings) that is not fully square. I too
> went for the linear bearing route and glued them
> in and it did solve the problem, but I am now
> looking for a way to move back to bushings as they
> are so much quieter. I think tightening the bolts
> of the frame after loosening the bushings and
> running it with the belt might be the *right*
> sequence of steps, anyway I'm sure your solution
> will also fix the problem.

Thanks Mike. I love the design of the Mendelmax but no matter how hard I stare at the bed I can't figure out what's exactly causing the problem. I unhooked the bushings and if they are tweaked they start to skip and this is amplified by the heavy aluminum bed. I thought about experimenting by filling the bed with structural foam and what not but it will be easy enough to wait for lm8uu or the 16mm bearings. I love the heated bed so hopefully the weight doesn't cause problems because I don't think I'd like using anything else after trying it.


kludgineer Wrote:
-------------------------------------------------------
> We have basically came to the same conclusion that
> the Y axis is over-constrained. It has a small
> amount of float built in to the Y shaft mount, but
> it is not enough to overcome the problems, and
> increasing the float causes enough looseness in
> the axis that it effects print quality. This has
> been tough to solve, because, apparently through
> sheer luck, my prototype of this Y axis worked
> brilliantly from day 1, with basically no binding
> or anything once I learned the trick of wiping the
> oil off the shafts. That lead me to an early
> assumption that the problems were simple alignment
> issues, and not a more fundamental design flaw. I
> apologize for that delay in comprehending the
> nature of the problem.
>
> We are in the process of re-working the Y axis
> mount to eliminate the over constraint. I have the
> design finished in my head, but I just need a few
> hours to sit down in front of the computer and do
> up the CAD models. Things have been insane for me
> for the last several months, but I have allocated
> time this weekend to complete that design, and I
> will get a few sets made up and and test it next
> week.
>
> Once we finish the design we will will offer a
> free upgrade for just the cost of postage to
> anyone who purchased a hardware kit or full kit
> from us. Barring unexpected problems with this
> design, I expect it to be available in about two
> weeks.
>
> -Kludgineer, AKA Maxbots

I finally got around to trying to fix this. I have been printing a lot on my mendelmax and the bed glides like butter but this shaking is still in the y axis that shows itself in the print. I can press down on the bed lightly and the 8mm smooth rods flex a lot so I think bigger rods are required since the bed is so heavy. Here's what I made so far and I'm going to bolt it in tomorrow to test it out. I do need something to insulate the bolts that will go into bed with some precision washers or spacers of some sort that I haven't found yet but that should be easy. I added a little bit of stock on each side vs what's in the drawing so the bolt hole areas would have more meat. I milled a pocket to try and save weight so the steppers wouldn't have to work so hard.







Edited 1 time(s). Last edit at 05/28/2012 02:38PM by LT1Pat.

I changed the table, stepper motor, pulley, and mounts. The only thing left is the belt and the tensioner and I'm beginning to realize the problem is with the zip tie tensioning system. I ordered a new longer belt to run rigidly mounted hold down tabs and a new y tensioner. I was pretty confident that the linear bearings and new y table would solve it but nope!