Frame stability, higher acceleration and crazy cooling

I'm building a large-format CoreXY homebrew. Build area is approximately a 48cm cube.

• ...
  RAMPS 1.4 (firmware may get changed or get re-written as needed). If performance is a problem, I'll move up
  DRV8825 stepper controllers with heat sinks and fans
  24V @ 100A max power (yes, almost 2.5kW), although this is definitely overkill.
  custom heated bed (my own design, since no one has a 50cm x 50cm version available)
  Bowden feeder on the frame to keep the movement light
  E3D v6 extruder (modifications below)
  device with operate in a slightly positive pressure inert gas envelope (carbon dioxide) for safety and consistency. I briefly considered N₂, but prefer CO₂ to avoid undetectable anoxia.
  Liquid cooling (H₂O, oil, or NH₃)
• ...

I've noted that frame torsion and flexing are much worse as you move the X axis and the tool tray at higher accelerations. I'm using 20x20 VSlot with Mini-wheels instead of steel rods because of previous problems with linear bearings wearing unevenly and excessive noise.

I'm targeting a typical movement speed of 2000mm/s, with an option of going faster if I can get filament down accurately. Motors are currently Nema23, but looking at moving up to Nema34.

To combat this stability problem in the frame, I've bolted the entire X-Y movement to a 1/4" thick aluminum plate and bolted the Z axis to the plate as well with some very sturdy hardware. I've noticed that the framework can "float" inside the enclosure on top of some vibration damping material to provide better isolation.
Why don't other people do it this way? Is there something I'm missing?

I'm in the process of re-engineering the hot end so I can push filament through fast enough for the extruder to keep up with the movement. It's mostly about lengthening the hot zone and keeping it stable thermally while maintaining thermal isolation outside the hot zone. Right now it looks like I'm tripling the length of the hot zone (new nozzles, multiple heating elements) and the best thermal insulation I can get. I'm planning on using an aerogel insulation wrapper (both for weight and thermal reasons). Has anyone done any robust cooling solutions with liquids?

I'm comfortable with liquid coolant because the thermal properties are so much better than air cooling, and I'm predicting some fierce hot-zone issues that forced air just can't solve. Besides, there's no added fan/housing weight, since the important "heavy" part of cooling (pump, heat exchanger) an not on the tool tray.

Poke holes in this please... I'm interested in all the things you might think I may have missed.

Please be aware that I have deliberately abandoned simplicity for it's own sake in the interests of reaching a specific performance goal: absurdly short print times.
Secondary goal is silent operation.

... are you serious with 2000mm/s? - most people are happy with reaching 200mm/s

---

Viktor
--------
Aufruf zum Projekt "Müll-freie Meere" - [reprap.org] -- Deutsche Facebook-Gruppe - [www.facebook.com]

Call for the project "garbage-free seas" - [reprap.org]

I'm used to industrial milling machines. Even 1 meter/sec is ambitious, but I want to stretch. I'm willing to go to Nema42 if necessary, but that puts me way outside the norm.

I already know that the Nema23 motors I'm using outperform the extruder/hot end right now, but I'm working to build a machine that prints between 10 and 100 times faster than a typical 3D printer. I'd prefer 100x, but I'm sure that's not yet possible. Maybe I can participate in the formation of a high speed, high precision printing revolution. Current print times of days make this technology unattractive to people with short attention spans (most of humanity), so improving that will greatly speed mass acceptance and adoption, (one of my tertiary goals).

I am aware of print speed issues for filament, and the easiest current solution is making 4-8 copies of whatever item I'm making to give the layers time to bond before I force the next layer on.

In direct contravention of the RepRap concept, I'm using a machine shop (CNC mill, lathe and surface grinder) for fabricating the parts I need. For the stresses involved in moving that fast, even short-cut carbon fiber impregnated filament doesn't have the strength characteristics I'm going to need. I may print some elements, but all the moving parts are essentially industrial movement.

This is NOT a tabletop machine. Between the cooling system and the gas cylinder (with computer-controlled regulator for the inert gas feed) it's more like a 30 cubic foot refrigerator format.

You need also to think about the hugh pressure in the nozzle at this fast speed. If you tried to print with a 0.25mm nozzle, this should go in the same direction.

The Ramps/Mega can handle a maximum of approx. 63,000 steps/s. With 2000mm/ you have just 31.5 steps/mm. I would prefer a Due for such big fast things as this is much faster and can go up to 320,000 steps/s with quadstepping.

---

Triffid Hunter's Calibration Guide

--> X <-- Drill for new Monitor

Most important Gcode.

... it's not about the pure speed ... but problems with extruding/dispensing fluid or semifluid materials through small nozzles and interconnecting of tracks/layers tends to grow quadratic with faster speeds

---

Viktor
--------
Aufruf zum Projekt "Müll-freie Meere" - [reprap.org] -- Deutsche Facebook-Gruppe - [www.facebook.com]

Call for the project "garbage-free seas" - [reprap.org]

There are several challenges to high speed printing, and I think you are focusing on the wrong ones.


I'll start with the mechanical limitations of a cartesian platform. Velocity is rarely an issue for 3d printers because they do not have enough time to reach maximum speed. Sliced 3d models generally have many short linear segments, and a zigzag infill pattern that requires many direction changes. The typical VMC would be lucky to reach 10% of its maximum velocity in a printer toolpath.

To put the problem in perspective, reaching a maximum speed of 2000mm/s from rest during a linear movement of 50mm requires an acceleration of 80,000mm/s^2. It is very difficult to properly convey just how large that number really is.

I'd be looking at 4060 extrusion and linear block/rail attached to something heavy. Even at modest accelerations 2020 starts to look like a wet noodle.


Some of your ideas seem extraneous - is the inert gas supposed to prevent fires? It will not stop the plastic from decomposing at high temperatures if that is what you are looking for. I also feel that liquid cooling can be sidestepped with proper design. Stratasys actually has some very clever ideas in this area. In simplest terms, you want only the extruder nozzle and bed to be in the heated chamber.



This eliminates the need for liquid cooling and is clean.


Building a large format printer that remains accurate while printing at even modest speeds is heavy and expensive. I would be careful about using a bowden. The concept becomes problematic on large format printers, and I'd be concerned that high extrusion speed/pressure will make it a poor choice.

I won't talk about the extruder since I have no experience there.

Quote
691175002
To put the problem in perspective, reaching a maximum speed of 2000mm/s from rest during a linear movement of 50mm requires an acceleration of 80,000mm/s^2. It is very difficult to properly convey just how large that number really is.

In terms of G-forces, it's about 8.15G, or roughly 1G less than a rocket bound for the ISS. Not impossible, but not reasonable.

As for extrusion, the extrusion has to keep pace with the head's movement, so for 2000mm/s you would have to be extruding that same amount, For simplicity's sake let's say you have a .3mm nozzle and a 3mm filament, meaning a 1/100 ratio for the feed to the output, that means you're still pushing the filament at 20mm/s through the hotend. The regular hotend is around 10 mm, meaning you would have to have enough heat input to melt the plastic in half of a second. Also, pressure becomes a major hurdle. I haven't done fluid dynamics in a while, but that might end up being something in the dozens of kPa, risking breaking an aluminum block of the (glowing hot) hot end even if the extruder doesn't strip from the hundreds of newtons of force it needs to push the filament that fast through the block.

To be honest, that block of specifications seems more like a buzzword pileup from a CNC catalog than a plan.

Quote
Feign
In terms of G-forces, it's about 8.15G, or roughly 1G less than a rocket bound for the ISS. Not impossible, but not reasonable.

Unreasonable is one way of putting it - I'm pretty sure that if you tried to accelerate a metal extruder at 8G the heat-break would snap in half. Especially with a long melt zone and multiple heater cartridges.

I haven't done the math but that kind of impulse is probably in line with dropping your printer from waist-height every single direction change.

Edited 1 time(s). Last edit at 01/09/2015 06:35PM by 691175002.

Your points about mass and acceleration are well taken, and have been on my mind for a while. I've been working on taking the mass of the moving X-Axis + tool tray + hot end to an absolute minimum to reduce that source of distortion caused by acceleration. I'm pretty sure I can keep the total moving mass under 400 grams so far, but the concern has been that once I start moving it fast, it just won't have enough structure to print at speed. I'm not crazy about giving up on the motion speed goal just yet, but the print quality or lack thereof will dictate how that goes. The alternative is to replicate a CNC milling machine's structure, and spending for really big steppers but that's an incredibly big scary machine (accelerating heavy stuff at 8-12g) that requires a buffer zone and ballistic shielding around the movement domain that won't fit in my garage.

The inert gas envelope is for chemical and fire safety. I've seen some pretty bad things happen to equipment that ran unattended (fire, grinding of gears into oblivion and smoking motors, metal fires and machines falling over), so if something goes wrong or even horribly wrong, it stays contained. I've got some curiosity regarding non-standard filament choices (including using PVC, other plastics that may emit toxic products and some non-thermoplastic materials), so I'm planning ahead. It's a short hop from an inert gas environment to vacuum, so if I get whacky and try using an electron beam for deposition it won't require a complete redesign, just shielding which I've already taken into account for energies up to 120keV).

As for the liquid cooling, I'm considering this as a solution to some of the heat problems I've calculated. Again, I'm starting with air, but preparing for the worst.

Why thank you, Feign! That's the nicest way of telling someone they're excrementally augmented to capacity I've heard since my 40's. I feel welcomed in the truest sense of the word...

I am a reasonable person with unreasonable aspirations. That's how discoveries often happen or, crashing and burning spectacularly, which is more likely.

Since it's a larger printer, I expect that there may be at least some instances where the tool tray could get moving pretty fast, since 480mm*sqrt(2)/2 is a little further than 50 mm.

I said tool tray because it doesn't have to strictly be a 3D printer. I could put a laser cutter, electron beam for ion deposition, water jet cutter, wire EDM, or a drive spindle for a micro-mill. Right now, it's a experimental 3D printer for thermoplastics. I have no idea what it will end as...

The pressures in the hot zone may require me to toss the Bowden in favor of a direct-drive component. 100kPa ends up being a little more than 14psi, which brings some interesting problems feeding and maintaining micron-level resolution for extruding and retracting. At that pressure it might be worthwhile to do the hot-end off the tray and insulate the heck out of a line bringing already-liquid plastic to the extrusion point. Could work, but maintaining thermal stability throughout the vastly longer flow path worries me. Even though you're unnecessarily insulting in your presentation, you do raise a good point. I'll try to focus on that contribution moving forward.

I'm sure I've missed something, even though I have just under 1000 pages of notes, design sketches, movement analysis, and preliminary calculations for everything from the dynamics to the wear characteristics of the materials and assemblies.

Peer review does work, assuming people are helpful.

Edited 1 time(s). Last edit at 01/10/2015 08:27AM by foamstone.

I think doing 10x reprap speed is do-able, but 100x is not. The acceleration is the problem. Cartesian printers just can't accelerate that fast! Delta printers can, though. I keep mine at 3000 mm/s^2, but with a few mods I can bring it to 1G.

Perhaps you should stick with your ultra-rigid machine design but limit it to accelerations that a fighter pilot could take. Then extrude out of a huge nozzle, and have a lathe-style rotary tool turret that switches to an endmill so you can cut material away. That's the easiest way to get the 100x build speed you want.

@Generic Default:

Very interesting! I like it!

I hadn't considered the idea of a hybrid machine that does both additive and subtractive composition. That's something I'll have to mull over some, because at that point laying down large volumes of material and then finish forming will get the time to completed part down a great deal...

So far, the problems I see with going to 100x print speed are still economically infeasible, but I'm still going to shoot for a 5x to 10x improvement. It looks like the main issue remaining is finding a way to accurately deposit filament at higher speed without giving up the precision.

Hotend design and machine rigidity are critical for what you want. The nozzle has to be exactly where you want, when you want, and any deviation from position or velocity will show up in print quality. Sharp corners will look rounded with overshoot. There is a thread that is relevant to this here;

thread here

Check out the part about "ringing ripples" as that will be a huge problem when you use higher accelerations. The key to avoiding it is to use a rigid setup, preferably low mass, and having very little backlash if any at all.

One way to keep detail and print fast is to have a small, detailed nozzle that does perimeters and then switch to a large nozzle for infill. Milling after large extrusions is another option. I'm early in the build process for a large delta printer that also does 5 axis milling, so I think we're after the same goal.

Good luck with whatever you come up with!

This thread is interesting... I just posed about some things about speed considerations a second back...

[forums.reprap.org]

Wow....2 m/s velocity within a 0.5 m cube?

That would need 80 m/s^2 acceleration just to crash the head at top speed into the far end stop.....ambitious is not the word I would use, but I would LOVE to see the test videos.


One thing I will add is that most of the cube designs I see posted around here lack one component I seemed to pick up a critical for rigidity in my limited mechanical engineering courses...namely crossbeams. Granted, many of the designs are likely sturdy enough without them for the forces of current printer designs, but for what you are talking about, I suggest you seriously consider some crossbeam placement....

hello everyone, i'm new here.that's some information i find with H-bot
here
2000mm/sec with 5G acceleration
repeatability +-0.05mm
positonal accuracy 0.8mm/meter


here
And there is the link printing with 750mm/s
and he said that he was printing with 0.1mm layer and scaledown 50%the cute octopus using ABS white
he was connecting the printer with computer and loading other program so there was some interrupt
and he didnt hv enough fan cooling,so he was printing failed on the overhang area



So,maybe you should targeting on 1000mm/s or improve the cooling and accurately 750mm/s . It will be more reality,after success then move on.

sorry for bad english. hopes it may help.

Edited 1 time(s). Last edit at 01/30/2015 02:42AM by nikolaos Y..

Quote
nikolaos Y.
here
And there is the link printing with 750mm/s

I'm just going to point out that he is not printing anywhere close to 750mm/s. He can type ten billion mm/s into the printer configuration for all I care, that doesn't mean its actually happening.

Just look at the video... 750mm/s would imply 30" of movement every second. If you watch the clock, that printer can't even hit 150mm/s. As I have been saying, top speed is meaningless when you are acceleration limited. Any idiot can achieve higher velocity - just use bigger timing pulleys. Acceleration requires $ and engineering.

I'm also almost 100% sure that H-Bot video is sub-2G as well. I'm sure they could build configurations that hit 5G, but video does not feature one.


If you want speed, you cannot start with a flimsy frame and trashy linear motion. This is how you do it properly:





I am only building this printer because I found some cheap listings. I have no intention of printing anywhere near the limits of these stages. Purchasing any of these parts new would cost $5000+,

@foamstone, I run a company (fusion3design.com) that designs and manufacturers 3D printers similar to what you're designing. Namely corexy drive system, large build volume, high print speeds, and bowden extruder system. Our swept volume is 306x306x306mm, which is a magnitude smaller than what you're attempting. I have a couple of comments based on my experience that may be helpful.

First, our peak travel speed is roughly 1m/s. This is with NEMA 17 motors and standard reprap electronics. However, the fastest print speeds we've been able to hit are around 250mm/s. The limiting factor turns out to be the rate at which you can feed and melt the plastic. Extrusion force and melt chamber pressure climb rapidly as feedrate increases. At some point you're going to a) strip the filament b) explode part of the bowden system (don't use PTC fittings) or c)or explode the hotend. I think this is going to be your main limiting factor. There are some excellent posts on the topic of feed rate limitations in the delta 3d printers google group. Hotend selection matters a lot. E3D has relatively high extrusion force for a given flow rate vs something like a jhead. Their new volcano option may help this, given the longer melt zone?

Second, bowden systems have hysteresis that results in excess plastic in corners and flow starvation in acceleration regions. In theory you can compensate for this with a simple first order approximation in the firmware but nobody's done it successfully yet on RepRap firmware. This, again, limits how fast you can print if you want to achieve accurate results.

In short, I think you're over-engineering the frame and motion mechanics. Your limiting factor is going to be the extruder/hotend system. Spending heaps of money and time on a motion system that can run at 2m/s, when you will be limited to 200-300mm/s print speeds (optimistically) doesn't make sense.

Quote
crispy1
Second, bowden systems have hysteresis that results in excess plastic in corners and flow starvation in acceleration regions. In theory you can compensate for this with a simple first order approximation in the firmware but nobody's done it successfully yet on RepRap firmware.

crispy1, I implemented Bowden extruder compensation in my fork of RepRapFirmware for Duet electronics. What I do is to make the extrusion amount E' related to the requested extrusion amount E using the formula E' = E + k dE/dt. In other words, advance the extruder an additional amount proportional to the rate of extrusion. The constant k is user-definable with a gcode (see [reprap.org]).

However, I have had limited success. With values of k up to 0.1, I don't see any improvement - there is still excess filament deposited where the direction changes. At larger values of k, it is necessary to retract filament during deceleration.With K = 0.2, this retraction causes the filament to get stuck in the Bowden tube of my Ormerod.

I have a couple more things to try. One is to reduce the amount of retraction I have configured in slic3r (currently 4mm), in case the problem is occurring when that 4mm retraction happens just after the retraction needed for Bowden compensation at the end of the previous printing move. Another is to try it out on my other printer, which is a Mini Kossel with E3d hot end.

Is this the sort of Bowden compensation you meant, or did you have a different algorithm in mind?

Edited 1 time(s). Last edit at 02/07/2015 04:00PM by dc42.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Quote
foamstone
...
RAMPS 1.4 (firmware may get changed or get re-written as needed). If performance is a problem, I'll move up
DRV8825 stepper controllers with heat sinks and fans
24V @ 100A max power (yes, almost 2.5kW), although this is definitely overkill.
...
Liquid cooling (H₂O, oil, or NH₃)

My 2p worth:

1. You obviously won't be building this on a budget, so ditch the RAMPS and get a modern 32-bit controller board, which can give you more precise motion, faster non-printing moves on a machine as big as the one you are planning, web interface etc. I'm a fan of the Duet, but the Smoothieboard is also popular.

2. 24V @ 100A - are you serious? Where are you going to get that PSU and how much will it cost? How thick will the cables need to be? It will be much easier and cheaper to use a mains-powered heated bed controlled by a zero-crossing SSR. You just need to think carefully about how to safely feed mains voltage to the bed heater. But you can get highly-flexible high voltage high current cable intended for test equipment, so it's not a big problem even with a moving bed (and even easier with the fixed bed of a delta).

3. Why liquid cooling? Unless you have a lot of heat to get rid of (e.g. because you have a Kraken hot end), surely this just adds complexity for no gain?

Edited 1 time(s). Last edit at 02/07/2015 05:01PM by dc42.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Quote
crispy1
The limiting factor turns out to be the rate at which you can feed and melt the plastic. Extrusion force and melt chamber pressure climb rapidly as feedrate increases. At some point you're going to a) strip the filament b) explode part of the bowden system (don't use PTC fittings) or c)or explode the hotend. I think this is going to be your main limiting factor.

How much does it help to increase the size of the orifice at the hotend?

Ok, strictly speaking, I love overbuilt homebrew machines, but for the sake of poking holes, you might want to re-evaluate your approach. Extruders and the mechanics of extruding plastic are a much bigger limiting factor than having a machine rigid enough to handle high speeds. I've heard of people printing at 400mm/s, but i'm dubious of anything past 200. Seems to hurt print quality more than anything, and restarting failed prints takes longer than running it at a reasonable speed the first time around.

If you're looking to drop print times, go for bigger nozzles. As some already mentioned, E3Ds Volcano would be a great option. You'd have to machine some new parts, but running a Volcano for infill and a detail nozzle for perimeters would be the way to go. (volcano sits about 9mm lower than the standard V6 hotends).

Depending on your budget, best bet would be shoot for a bit more reasonable speed, spend a little less on frame components and upgrade the guts. Smoothie board or something from panucatt for the electronics, Chimera or a Kraken for the hot end. Kraken is only 94g with all 4 nozzles...food for thought!

just to mention an idea I have for a while now,and it might be funny and dumb, but I think it's the only way to really print fast and big:

figure a sheet of plastic with a preset thickness, this sheet it's in an oven/conveyor system, a laser cuts the shape you want and then the sheet its heated to temperature, once ready the sheet is moved to the build plate and press against it. same process layer after layer.

Good things?

1. lasers can be stationary and can cut really fast and precise.
2. plastic sheet are pretty easy to acquire.
3. press on layers will reduce delamination.

Bad things

1. A lot of waste material
2. Only can work in an Industrial setting
3. REALLY expensive.
4. Will be a waste to print only one small piece

The technology you said is call LOM(Laminated object manufacturing)
wiki

i had reseach about the main problem few weeks,
the main problem of the limitation is the X,Yaxis and the extruder

if you want a high accurate x.y axis with lower price,
using a made in china laser cuts is just around 247.3USD chinese ebay
and adding the z axis to become a 3d printer
OpenSLS
if you wanna make a fast print 3d printer,FDM is not a good choose because the motor and the price is not proportional ,without consider the synchronous speed of the extruder

So,why dont use other technic like SLS or SLA,they have no XYaxis and extruder,and some SLS can be print around 10m/s , 100 times speed comparison with fdm (100mm/s)
if you using several laser head ,you can even speed up 2-5times.
the sls powder is even cheaper than FDM filament roll

nikolaos Y.,

Thanks for the info....

So my idea isn't as crazy as I thought......

But someone beat me in creating it........, oh well

I like the ambition in your project. I'm also testing out different ideas with the hope of improving speed and accuracy in 3d printers, and I agree that sometimes you need to be a bit crazy with your goals and methods to actually get results (scientific breakthroughs a good example).

Some of your goals are probably a bit too enthusiastic, and you'll run into engineering problems ranging from plastic flow rates to near impossibility of achieving the acceleration you need, as well as many others beyond doubt. It may be worth considering making your platform modular enough to test different ideas once you realize that one doesn't work (e.g. no way you're getting 100x or even 10x reprap speeds on the first iteration). That said, I am very interested to see how your project evolves with time if you decide to stick to it.