X Y axis accuracy doubled.

How about using chains instead of belts?

Nice one Madkite. You have halved the gear ratio on your stepper motors. I am currently designing a gear box to lower the ratio of my X and Y motors to see if I get more accurate prints. I mite adopt your idea as well and have even more accuracy ! I'm hoping to get them as accurate as my Z axis @.@ ....... he he well see.

With 1/16 stepper controllers, we can already get 80 steps/mm or 0.0125mm with the usual belt design. I don't think the x-y is limiting the resolution. If you want to improve it further, you should look at the extruder.

Thanks Spenc.

brnrd. The benefit's are not just in accuracy but also in the reduction of torque required by the motor.

Having tested the accuracy quite thoroughly. The extruder system available are very accurate. The thread consistency is almost perfect and so if you have all the settings correct you can easily reproduce the accuracy to the limits of the X Y axis. The tricky part is on internal curved surfaces. But these are all code adjustments to allow for the way the plastic is deposited. Having smaller nozzles does not make for more accurate prints but more detailed. The consistency of the feed and depositing is what makes it accurate and it is already very good. I'm not convinced that you can push the extruder to go much faster than some people have already. The speed is also dependent on what material you are printing. I'm actually trying to invent a new depositing system to overcome the limitations of the current system and allow for a much wider rage on materials to be printed not only with the same head but in the same print. Hopefully if I can get the time, inspiration and money (Don't really need much money as its designed to be cheep.) I will try to make one and make it work over winter. A proof of concept at least anyway.

I see no reason why having a higher accuracy is not desirable if it can be done without needing to introduce high precision expensive systems and components.

You could use chains but the weight might limit the system speed. It would be super strong and last for ages though. Not a bad idea for a very large machine perhaps.

I can also report that the Kevlar belt I was testing earlier in this thread has still not failed and worked perfectly without adjustment. Before anyone asks all the likes and explanations are in an earlier posts.

I would highly recommend the small extra investment in them.

---

Make your Mendel twice as accurate.
[www.thingiverse.com]

I have already swapped over to metal bead chain for my Y and Z axis, but i'm thinking of changing that again to chain driven. I have a heavy tiled build plate on my Mondo which is one of the largest of the RepRaps :/.

Madkite: I think everyone agrees that the concept looks great but on the claim of increased accuracy can you show any proof of that your parts are more accurate?

On such claims It would be real interesting to post the STL for a test part and let anyone who want to print the part and publish dimensions. Accuracy to me is dimensions from top to bottom, squarness, ID, ect. Could be something like the 20mm test cube with a hole running throught it. If accuracy is important I'm sure you have a micrometer. or caliper? Folks could even post a photo.

This thread has gone on for two pages, it would be interesting to see if it really works.

Quote
Madkite
Having tested the accuracy quite thoroughly. The extruder system available are very accurate. The thread consistency is almost perfect and so if you have all the settings correct you can easily reproduce the accuracy to the limits of the X Y axis.

I think you can only make this argument if you never have to start and stop the extruder during the print. Most of the imperfections occur at the start and end of each thread. In my view, this is the major obstacle to getting high quality prints. There's already sufficient torque from the NEMA 17 motors running at 1/16 steps to drive most versions of the Mendel.

I think you're confusing resolution with accuracy. The resolution is increased from 0.0125 mm per step to 0.00625 mm per step. But this is not the accuracy. The step angle accuracy of a NEMA 17 is typically +/-5% of a 1.8 degree step. Assuming the same applies for 1/16 steps, the accuracy in the Mendel would be 0.000625 mm. By decreasing the step size with your belt system, you have improved the accuracy to 0.0003125 mm. I don't think this improvement is noticeable with any Mendel.

Quote

The step angle accuracy of a NEMA 17 is typically +/-5% of a 1.8 degree step.

One thing you've forgotten here is, the geometry of a part isn't related to the geometry of the steps. Some specific geometry could require 100.49 steps, so the closest you can do is to move either 100.00 steps or 101.00 steps, both times missing the target by about 0.5 steps. This 5% figure might be right, but doesn't really matter.

Standard Mendels with 1/16 microstepping have currently about 80 steps/mm, so the (theoretical) accuracy is 0.0125 mm. The system proposed here doubles this to 0.00625 mm.

That said, I think a Mendel frame is waaays to wobbly to make this kind of accuracy realistic. The advantage of this proposal isn't increased step resolution, but reduction of forces. With half the force required to accelerate the extruder, the frame bends only half as much when printing - approximately.

---

Generation 7 Electronics	Teacup Firmware	RepRap DIY

Quote

With half the force required to accelerate the extruder, the frame bends only half as much when printing - approximately.

Eh? f = ma With m and a the same, surely f is the same.

---

[www.hydraraptor.blogspot.com]

Actually since the max speed is half what it would be without the mod the a(mod) = a(nomod)/2 so f(mod) = f(nomod)/2.

---

Bob Morrison
Wörth am Rhein, Germany
"Luke, use the source!"
BLOG - PHOTOS - Thingiverse

brnrd Wrote:
-------------------------------------------------------

> I think you're confusing resolution with accuracy.
> The resolution is increased from 0.0125 mm per
> step to 0.00625 mm per step. But this is not the
> accuracy. The step angle accuracy of a NEMA 17 is
> typically +/-5% of a 1.8 degree step. Assuming the
> same applies for 1/16 steps, the accuracy in the
> Mendel would be 0.000625 mm.

lets agree that:

Resolution = how detailed u are able get your prints e.g. Software,firmware,Steps/mm,Tip size etc etc

Accuracy = how close u can get your printer to print out to your resolution in real life e.g. Belt tension, frame rigidness, motor torque/(Current), vibration. etc

Lowering your gearing ratio does lower your Stepspermm (possibly to ridiculously low amounts) but it also lowers your margin of error on how close u can get to your resolution, there for increasing your accuracy. As long as your belts have sufficient tension of course. Bring on chain drive rapraps !

Quote
Traumflug
One thing you've forgotten here is, the geometry of a part isn't related to the geometry of the steps. Some specific geometry could require 100.49 steps, so the closest you can do is to move either 100.00 steps or 101.00 steps, both times missing the target by about 0.5 steps. This 5% figure might be right, but doesn't really matter.

Standard Mendels with 1/16 microstepping have currently about 80 steps/mm, so the (theoretical) accuracy is 0.0125 mm. The system proposed here doubles this to 0.00625 mm.

This is called resolution, not accuracy. Here's a definition from merriam-webster:

Quote

a : the process or capability of making distinguishable the individual parts of an object, closely adjacent optical images, or sources of light
b : a measure of the sharpness of an image or of the fineness with which a device (as a video display, printer, or scanner) can produce or record such an image usually expressed as the total number or density of pixels in the image [a resolution of 1200 dots per inch]

The step size is basically equivalent to the pixel size in a monitor or printer except since the Mendel is vector based instead of being pixel based, it is probably more like the old digital pen plotters.

And here is the definition of resolution for 3D printers from Wikipedia:

Quote
wikipedia
Resolution is given in layer thickness and X-Y resolution in dpi. Typical layer thickness is around 100 micrometres (0.1 mm), although some machines such as the Objet Connex can print layers as thin as 16 micrometres.[9] X-Y resolution is comparable to that of laser printers. The particles (3D dots) are around 50 to 100 micrometres (0.05-0.1 mm) in diameter.

But I also don't think that this gain in resolution would be noticeable in the Mendel.

I will make a video of a test part and how I configure this week if I can. I'm a bit busy at the moment.

I have confirmed this accuracy with calipers and I'm used to working very accurately as I do machine work very often.

It is also necessary to configure your machine to its actual output not a theoretical number of steps. One of my axis has 81.something instead of the theoretical 80. This could be a pulley error or any number of other things. Its not belt slack or you get other printing errors. But that tiny error makes a 1mm error on my test cube. Once you correct for these mechanical inaccuracies I found it easy to reproduce the accuracy to a much higher degree than I thought I could.

I can only get 40 steps with my set up normally so it may be less of a problem with other electronics but I did also find this mod almost completely eliminated backlash which I had a problem with before. A lot of the increase in accuracy I have noticed may be to do with the reduction of stress on key parts of the system rather than just because the steps have increased.

The thread may start with a blob or other small inconsistency but that is just something to knock off after its printed. The main movement around a perimeter should be completely consistent and smooth. And if its consistent than it can be done accurately.

---

Make your Mendel twice as accurate.
[www.thingiverse.com]

One approach would be to print about 10 20mm test cubes. But there really should be more discussion around what to test print...

Measure the height and widths in random locations. Meaning over different places on the face. Try not to bias the experiment by remembering what it was the last time. Possibly measure them as they are printed, not all at once.

Calculate a mean and standard deviation for the two columns of numbers. The mean represents how accurate your machine is to printing to 20mm. The standard deviation represents how repeatabile your machine is. Machines can be accurate but not repeatable, which is bad, and repeatable but not accurate, which is much easier to fix. In the end we really care how accurate the machine is? Shouldn't accuracy should be in the parts?

If we had some sort of a spec like 20 +-0.2mm you could generate quality statistics like Cp (the ratio of the variation of the design to your variation) or Cpk which includes centering. You can look up the math.

So a Cpk for height and a Cpk for width would be great assessment of accuracy for a very small study of 10 printed parts.

I know the errors in accuracy are repeatable. I simply wouldn't be able to correct them and see the correction I enter represented in the next test precisely in relation to the correction. I'm very familiar with how to measure. I make machine parts regularly so I'm sued to seeing if things are really parallel or flat. I get far more error from shrinking forces creating stress than anything else. I probably wouldn't suffer as much if I were printing in PLA instead of ABS.

The height on my machine is no more accurate than any other. I didn't worry about modifying it as its way more accurate than all the other movements anyway. I can test that for setting but it is not telling anybody here anything new about the machine as its standard.

---

Make your Mendel twice as accurate.
[www.thingiverse.com]

Should be able to make a video Sunday if evrything goes to plan.

So can do my normal test 80mm by 80mm print and calibrate evrything as it need doing now. Any test object you want me to print pleas ask. Nothing ridiculously big though. And I'm not sure I have inside curves accurate yet so I'm only looking for circularity as that's a code setting to get the diameter right.

Once I have it all done I will put the test object up and the video. Though it may be a few days later depending on my time.

---

Make your Mendel twice as accurate.
[www.thingiverse.com]

haa I didn't seen that the thread had a second page, but wanted to reply mostly because today I finally been able to print with my VertX mod that integrate this idea
[reprap.org]

I made no calculations or else at the base of this choice, just liked how it moves and wanted to try it.
Will follow attentively this part of the discussion (and/or may post also about test prints)

Edited 1 time(s). Last edit at 11/10/2011 05:38PM by Emmanuel.

Very nice design Emmanuel. With your Vert X axis configuration, what what is your build height?

How about a test on something more interesting that a cube?

Cable Chain by RotoScan

Specifically the Cablechainx4 version. This part has many features that can be measured with a calipers like length, width, thickness, a bridge, two round posts, the flange that makes up the side walls, and what makes this particulary interesting is that the four parts are printed in different orientations. So any differences between the X&Y axis can be see across all parts of the same build. What makes this also interesting is that you could also test the interaction of size and strength, by measuring at what force the chain comes apart when pulled. I guess the logic is that poorly made parts will disconnect easier, which may or may not be true, but its always fun to break stuff.

What do you think?

Been calibrating and taking video. Got in a little muddle as I got the X and Y muddled up (Don't do calibrating hung over!) but more shockingly. My Y axis is really worn. Got a nice flat on the rod so its not running quite straight anymore.

So I'm going to finish my video and when I don't need to print any more I will design a linear baring mount to go on the Y axis and change the rods.

---

Make your Mendel twice as accurate.
[www.thingiverse.com]

It's due to one or more of the bearings not being perfectly straight on the rods so it rubs a little. In the meantime, you can rotate your rod so that the bearings are riding on the good parts of the rod. I've had to do this with my X axis a couple of times or so before changing the rods and then changing the x carriage. I've also applied some light silicon oil coating on the rods to reduce wear. From what I've been hearing, the linear bearings tend to also cut into the rods after a while. Perhaps this depends on the quality of the linear bearings and lubrication.

The plastics I got in the original Kit were less than precise. Far worse then the ones i have printed. The X axis has a carriage I printed myself and its perfect. But the Y has one of the originals and I don't need to measure it to know its at an angle. My plastics that I have in the sets I'm printing are probably ok but I'm just sick of the silly adjustment system that you cant get at.

I will be getting high quality bearings for it. Still I managed to get the X and Y set too within .05mm of an 80mm test today. But I won't be able to do better than that until I fix this.

Going to print 20mm test cubes tomorrow. Doing a set of 3 where one is square to the axis and the others at odd angles.

These things always take longer than I anticipate.

---

Make your Mendel twice as accurate.
[www.thingiverse.com]

Just a quick update. One of my wire wound heater resistors has exploded so I am having to wait for a replacement but apart from that does anyone have any ideas as to why the size should be too large on smaller objects but correct for bigger objects. Is this a skeinforge settings problem or a steps per mm problem? I'm talking X Y here only. I'm not talking very large diferences. 0.2mm at the most. But I can't work out where the problem is yet.

---

Make your Mendel twice as accurate.
[www.thingiverse.com]

The size of an object has two components: the distance the axis moves and the width of the filament. If your flow rate is a little high and your steps per mm a little low you could have small objects too big but larger ones correct. This would happen if you calibrate from measuring a large object when your flow rate is not correct.

---

[www.hydraraptor.blogspot.com]

That's what I had been thinking might be the problem. It was overfilling slightly and I was setting it when it literally went bang. Going to be a little while before I can do any tests now though. Bought some higher power resistors which should take any accidental abuse and last a long time. will take some video of that too.

Going to be a long video at this rate. Nothings ever simple.

---

Make your Mendel twice as accurate.
[www.thingiverse.com]

Madkite Wrote:
-------------------------------------------------------
> Just a quick update. One of my wire wound heater
> resistors has exploded so I am having to wait for
>...

How violent was the explosion? Do we have to worry about pieces of ceramics fying into us? I'm wondering how much of a safety issue this is. I did a google search and I found this video. After watching it, I will now make sure that my face is not in line with the hole for the vitreous resistor when watching the printer. Right now, the hole faces towards the front and back. I should probably rotate the hot end (Budaschnozzle) 90 degrees.

Quote
brnrd
Very nice design Emmanuel. With your Vert X axis
configuration, what what is your build height?

In theory I have the same build height, or maybe even a little more but that depends also of the hotend, and the v6 is quite long.
But I mostly forget the fan, and it bump on the wooden plate that hold my gen6 (at z = 60-70mm maybe ^^'), so I'll try to mount it on a side of the machine instead.

It just went pop and one end come out with quite a bit of smoke. Hardly anything to worry about. But smelly. You'll see in the video.

---

Make your Mendel twice as accurate.
[www.thingiverse.com]

I have blown a few wirewound resistors, but no flying metal. I think because they are made for high power, they die in a controlled way. The smell is probably the most dangerous, it's pretty bad.

---

---

reprapworld.com
PLA, ABS, PETG multiple colors (3 and 1.75 mm) €16,49 / kg
Megatronics 3.0 €79.99 / Minitronics €37,19

Now that this has been around for a few more months, how is the Kevlar belt holding up?

---

www.Fablicator.com