EEVblog #257 - Makerbot Troubleshooting

[BravoV]

Dave, after following you & Phil for a very long hours since the beginning from unpacking, live streaming both on assembling and the latest 1st test run, and with all the troubles that both of you went through, yet we haven't seen the final satisfactory result.

With the current price level, also with what both of you have been experiencing so far, would you recommend this say to your best friend to buy this thing ?

Really like to hear your opinion.

[EEVblog]

Dave, after following you & Phil for a very long hours since the beginning from unpacking, live streaming both on assembling and the latest 1st test run, and with all the troubles that both of you went through, yet we haven't seen the final satisfactory result.
With the current price level, also with what both of you have been experiencing so far, would you recommend this say to your best friend to buy this thing ?
Really like to hear your opinion.

The kit, no. But that's moot now because the Replicator is sold fully built and tested only.
I'm a bit concerned about the apparent intolerance in the design, it does seem rather touchy. Whether or not that's the same for the new Replicator, I do not know, you'd have to ask someone who owns the replicator. Makerbot have said they'll eventually send me one, so we'll see.
But it does now work, pretty darn well in fact, it's just touchy and we need to tweak the values for complex designs. That's not something should need to be done however IMO,  it should either be tolerant of such things, or should handle it automatically.
Or perhaps we haven't fully set it up properly yet, in which case I don't know why, we spent a long time following all the instructions.

I don't know if similar commercial models are this fussy.

Dave.

[firewalker]

I hadn't seen a model build by such a machine until now. I was under the impression that you could use it to print "useful stuff" (like gears etc).  Seeing the cylinder, for example, I think you can only print chess pieces like objects. I am guessing that the way for serious 3D printing is a polymeric Z printer.

Alexander.

[EEVblog]

I hadn't seen a model build by such a machine until now. I was under the impression that you could use it to print "useful stuff" (like gears etc).  Seeing the cylinder, for example, I think you can only print chess pieces like objects. I am guessing that the way for serious 3D printing is a polymeric Z printer.

You can print anything you want, like gears. The cylinder was just an example.

Dave.

[firewalker]

I hadn't seen a model build by such a machine until now. I was under the impression that you could use it to print "useful stuff" (like gears etc).  Seeing the cylinder, for example, I think you can only print chess pieces like objects. I am guessing that the way for serious 3D printing is a polymeric Z printer.

You can print anything you want, like gears. The cylinder was just an example.

Dave.

Off curse you can. I don't know if it will be usable. I think the integration between the layers of the ABS will be an issue for moving parts or for small pieces.

E.g.



https://www.eevblog.com/forum/general-chat/company-for-replicating-small-plastic-parts/

[ecat]

...
But it does now work, pretty darn well in fact, it's just touchy and we need to tweak the values for complex designs. That's not something should need to be done however IMO,  it should either be tolerant of such things, or should handle it automatically.
...
Dave.

Put simply, accurate and repeatable linear motion is a bitch. This may change once room temperature super-conductors reach a reasonable price point but until that day...

Alignment, manufacturing tolerance, wear, stretch, elasticity, load, dimensional changes due to temperature all conspire against accurate and repeatable linear positioning. You've already encountered an alignment issue where having one pulley higher than the other causes what should be a 10mm movement in the horizontal to occur along some hypotenuse. A quick tweak and the heights are close enough, but what is 'close enough'? I mean elasticity of the belt drive? Changes due to temperature? These numbers are smaller than a hair on a gnats bollocks.

To achieve a resolution of 0.1mm, the average width of a strand of human hair, you need relative accuracy and repeatability better than 0.05mm and ideally closer to 0.01mm, approximately half the thickness of a cigarette paper. Few ever think of measuring down to the nearest 1/100 of a mm and while it is one thing to achieve this level of repeatability, maintaining it over a period of time is totally different ball game.

I imagine the source of greatest error in the Makerbot will be due to the tension, or lack there of, in the belt. Any slack in one side of the belt will be transferred to the other side of the belt each and every time the stepper motor changes direction - a major contributor to backlash. The belt will always be subject to stretching, especially when the belt is new. The belt will not stretch or wear evenly, especially if all printing takes place near the middle of the table introducing the unwelcome spectre of variation along the length of the belt.

When thinking about stretching don't focus on the mass of the moving parts, focus instead on the forces involved. Accelerating from 0mm/s to 30mm/s in the space 0.1mm requires a force F, accelerating from 0mm to 300mm/s in the space of 0.1mm requires ten times the force if I remember my school book physics. Every change in direction requires F to bring the moving mass to a dead stop and F to accelerate back along the opposite path. They all add up over time and all the little Fs are out to get you.

Heh, that's just some of the issues with the belt covered, hopefully it gives you some idea of the issues surrounding  accurate and repeatable linear motion and the dimensions of scale you are trying to achieve. From some of the examples I've seen I think these 3D printers are capable of incredible results but you're going to need several £100s of precision ballscrews to approach long term reliablity and repeatability and even then wear, temperature, elasticity,  manufacturing tolerance...

Accurate and repeatable linear motion is a bitch!

[EEVblog]

Accurate and repeatable linear motion is a bitch!

It just needs absolute position feedback.
A $10 pair of digital calipers is incredibly accurate and repeatable for example.

Dave.

[Psi]

It just needs absolute position feedback.

i was a bit surprised it didn't have that already

[ecat]

Accurate and repeatable linear motion is a bitch!

It just needs absolute position feedback.
A $10 pair of digital calipers is incredibly accurate and repeatable for example.

Dave.

The $10 callipers are still subject to all the wear, manufacturing tolerance, elasticity, temperature etc issues that plague the actual machine giving you another whole set of unknowns to ponder.

Quick experiment: Take a pair of callipers apart and you'll find a little metal bar that can be adjusted to keep the sliding part true to the long edge. Now pick a short range to travel over, say 20mm and slide over that range 10,000 times - in Makerbot terms this is equivalent to 100 movements per layer by 100 layers, 10mm high at 0.1mm resolution which sounds reasonable.  Now tighten up the little bar if necessary and repeat the procedure to 'make' 50 similar parts. You are now at 500,000 transitions over the selected 20mm range, tighten up the little bar again and try to move the head over the full measurement range.

My guess is the head will stick which leaves you with a choice of replacing the callipers or loosening the bar and accepting the sloppy measurement at some part of the scale. In the latter case the positional feedback is screwed, in the former case you have a new set of callipers but the machine itself has been subject to the same number of potentially wear inducing transitions as the original callipers and may itself be exhibiting signs of slop which will give your feedback system a headache.

Manual metal working machines can be fitted with positional feedback systems, these usually use some sort of glass rod to sense position and that's all I know about them. Of course manual machines are seldom subject to the high number of transitions we expect from their computer controlled equivalents, that is manual machines are subject to far less wear. Even so, an accurate scale is of little use on a machine which is out of trim.

There may well be some computer controlled machinery that does use absolute feedback but from my experience(1) accuracy and repeatability are more commonly addressed by providing versatile calibration options. A good example is a screw map - I don't know the proper name but screw map sounds good to me - an accurate measurement device (calibration transfer standard?) is driven by the linear system and a map is made relating machine position to actual position, once complete the machine runs sans feedback until the next calibration is due, maybe some tweaks are made every xxx hours or so and the linear system is replaced once slop becomes an issue. The calibration standard is portable and subject to little wear and more importantly the machine is in a state known to be good enough for the job at hand.

Anyway, witter, witter. In short, much like divining for oil, if absolute positional feedback were a good idea industry would use it more often, could save a fortune on quality control.


(1) Which is limited at best, but this is the internet, a place much like America, a place where opinion, superstition and apocryphal tales carry more import than fact