We have FDM and we make FDM machines.

(Selvakumaran Ganesan) #1

We have FDM and we make FDM machines. Today one of a customer asked me will these FDM can be used for making Engineering grade prototypes? He wanted to print parts and wanted to fit as it is without post-processing. I was stunned and unable to answer properly.

I wanted 3d printing experts opinion on building a higher grade FDM machine.

We use Hiwin linear guides, Wantai motors, Powge GT2 belts and pulleys, E3d original hot end, MKS Gen V1.4 boards, with CoreXY structure (4mm Aluminium for Top CoreXY & for X carriage) with Marlin, BL Touch bed sensor etc.

What are the components we can change or we can use to get those higher tolerance levels?

I will be ready for the price to get the higher levels of accuracy.

Thanks in advance.

(George Novtekov) #2

You need to define more what engineering grade mean for you client. If he expect parts that are with tide tolerances like 10 microns I believe there is nothing in the opensource FDM space capable to deliver this accuracy. For suck a tolerances he need combined CNC and FDM. IF he want engineering grade plastics to be processed with improved quality that the normal printers I would suggest you to improve this:

  • Get a rid of MKS and switch to 32bit board with good drivers from TMC range. and pair them with motors that do not resonate that much.
  • You may also decide to go for 9mm wide belt for reducing the ringing.
  • Get high quality pulleys.
  • Get good ball screw on Z axis.
  • Make it rigid and heavy.

Avoid my mistakes:
I went for the best mechanics and I found hard way that open source software is not there yet. Marlin and Repetier got something that developers called Jerk but this is not and simple acceleration ramp. This work for belts as they are springy but when you go for rigid and high accurate mechanics and this poor motion algorithms this start to look ugly. My printer is made from integrated blocks ballscrew and 4 point of contact linear bearing in single rail (THK KR33 modules). When I used suggested Jerk and Acceleration the sound is like nailing … Basically you can’t assume on rigid mechanisms that you can’t change the direction just like this if the speed is less than XXX. When you try to remove the jerk other things start to happen like planner can’t plan acceleration correctly when you have circle with too many segments. Also you have significant period for acceleration and deceleration which create no linear extrusion that is a problem but Pressure Advance helps a lot there.

(George Novtekov) #3

Forgot to mention :
-Go for direct drive extruders like Bondtech.
-Go for 0.9 degree steppers as they overshoot lower than standard 1.8.

(Randy Dierks) #4

1 - ballscrews on all axis, preferably precision ground ballscrews.
2 - replace all stepper motors with servo motors and encoders.
3 - use cnc software designed for the closed loop, servo motor based system.
4 - increase mass and rigidity of printers frame.

It would be slower, but offer much higher precision, and more importantly, better repeatability.

I am currently building an all ballscrew, servo motor, based cnc router, that I intend to also mount a hotend on to give this a try. So we will eventually see if the improvement is significant enough to justify the added expense.

(MidnightVisions) #5

This reads like advertising, delete.

(ThantiK) #6

@MidnightVisions it’s not. Check the past submission history.

(ThantiK) #7

@Selvakumaran_Ganesan , honestly - it comes down to your slicer and firmware. Use a professional grade slicer like Simplify3D and make sure all of your firmware settings are tuned perfectly for your machine. You should be able to make engineering-grade parts come out of that machine so long as they follow some basic principles of 3D printing.

(George Novtekov) #8

@Randy_Dierks have you ever built printer with ballscrews to recommend it? It is much worse than belts so if you don’t tried don’t assume its good. I already build mine and I deeply regred about my choise.missing/deleted image from Google+

(Randy Dierks) #9

@George_Novtekov no I have not. I do have a full machine shop though and I am very familiar with what is required to make ballscrews work. Specifically, the machine, and complete drive train must be very rigid. Additionally, while I have done milling machine conversions using both stepper and servo motors, in the future I would prefer servos with encoders as they are much more reliably accurate. That said, it may not work, but just because it wasn’t successful for you or I, does not mean that it will not work. The fact is that in the precision machining industry, precision ground ballscrews are king.

(George Novtekov) #10

@ThantiK I bough and used Simplify3d for years but recently I tried Cura and seems that recently S3D left behind. Cura manage the extrusion flow much better than S3d and the result are much better as Cura msnage to remove zits completely in combination with pressure advance. In contrast S3d I don’t now why but don’t perform well with pressure advance. So despite I have spend money for S3d I am considering to switch to Cura as results are really better.

(George Novtekov) #11

@Randy_Dierks So do I and thats why I did it. But this is not how printer work! Thats why don’t recommend something that you didn’t tried it. Balscrew don’t like the idea to instantly change the direction when you go bellow 600mm/s. Sadly thid is how the Marlin and Repetier work. And just for your information my printer eorks good but in order to operate without dhaking and nailing during every change in direction I sm using jerk of 200mm/min and acceleration of 280mm/s2 which is like 5 to 10 times lower than typical user. The best part is that if you use 9mm gates GT3 belt and gates pulleys you will have same quality as ballscrews. Possition accuracy under 10 microns because you don’t have significant forces on the extruder and you can use CoreXY.

(Jeff DeMaagd) #12

It seems the biggest reason belts get a bad rap is hobbyists only experience the knock-offs. Well-selected belts on a properly built and tuned machine will give you good results.

(Jeff DeMaagd) #13

There’s ways to improve the process but I wonder if the request was even realistic. Heck, I don’t know of a manufacturing process that takes a raw material and produces a perfectly finished, ready to use part once it’s done with no extra work. You’re talking about machines that melt a material and push it through a hole and expecting a perfect part every time and no extra work once the machine is done. Plastic shrinks a bit as it cools too, much like a casting and it can shrink in unexpected ways, and the resources to predict and compensate would currently be enormous.

(ThantiK) #14

@Jeff_DeMaagd CNC machining does this every day. It may not produce ready to use machines but it certainly produces 100% finished, ready to use parts on the regular.

(Jeff DeMaagd) #15

I considered that but it’s usually more complicated than that. Especially if you consider additional stages, tumbling or coatings (painted, plated, passivated or anodized) or other treatments and inspections.

(Randy Dierks) #16

@George_Novtekov interesting, by the way, that is a very nice looking machine. Just comparing to your machine in the picture, mine is going to be much more massive. I am using 20 mm ballscrews, driven by servo motors. It will be interesting to see if I run into the same issue.

(ThantiK) #17

@Jeff_DeMaagd Delrin/Acetal. Not all CNC’d parts are metal or need additional coatings, etc.

(Jeff DeMaagd) #18

K. All one step, single operation? No revising, refixturing, or rechucking?

(ThantiK) #19

With a 5-6 axis, sure. All of that can be a single process. These machines regularly rotate parts in/out of chucks without human intervention.

(Jeff DeMaagd) #20

Well then that puts it into yet another class of cost and complexity doesn’t it? Illustrating how tall of an order it is.