GUS v2 Goal Setting/Discussion

GUS v2 Goal Setting/Discussion

Postby Nicholas Seward » Thu Aug 21, 2014 9:49 pm

I have talked and hinted about this in many different threads all over the internet. I have to admit that I have a pretty well defined design waiting to emerge from my brain. I have gone through 100s of mental iterations and have converged on the plan I will line out below. All I need is time. However, time and time again, reader suggestions have drastically changed the direction of my designs so this is a very valuable.

What is wrong with v1? (I will update this as you guys come up with more.)
  1. The prints require a large parent printer. (200x200 is nice)
  2. The sliding of string on the bolts is not ideal.
  3. The cantilevered bolts can bend over time.
  4. The gears should probably be printed with their axis pointing in the z direction.
  5. Calibration and leveling is a challenge.
  6. Printed spools should be trued or machined from the start.
  7. Stringing requires you to be a zen master.
  8. The spring add some hysteresis.
  9. Steppers can skip steps just from the string friction.
  10. The basalt bed is awesome but it trips people up so the official design should use round glass.
  11. Bowden only (I am not saying that bowdens aren't great but why not make it either or so I can bypass the whole debate.)
  12. The hub doesn't have room for much experimentation.
  13. Firmware
  14. Requires lots of plastic.
  15. All the joints are one sided.
  16. No cool LEDs
  17. No cooling fan.

Fixes
  1. I will split the gear from the arm.
  2. I will use 625vv V groove bearings.
  3. Don't cantilever.
  4. Now that the gears are separate just print them correctly.
  5. FSRs will be located in the hub.
  6. Buy machined spools.
  7. The new arms have much more access for stringing. With the separate gears there is lots of finger room.
  8. The XY printed gears should be pretty round and the string should be long enough that no spring should be needed.
  9. The 625vv bearings will fix this.
  10. I will use round glass that is held down with binder clips. (fancy)
  11. The hub will be expanded to be about 100mm in diameter. I will print my own lazy susan and use acetal balls. This will give plenty of room for a stepper to sit on top.
  12. The lazy susan hub will have a generic mount plat that will allow for any number of different hotends and coldends.
  13. Do it. (It is mostly done except I am going to have a different shoulder offset vs hub offset.)
  14. Too much plastic isn't really a problem. I will do what I can but rigidity is more important.
  15. I will switch to double shear joints. This by itself will my the effector more rigid.
  16. I am designing the bottom of the lazy susan to accept a LED ring.
  17. I will duct an axial fan from the top.

Additional Goals Collected from Below
  • Provide paper templates for cutting the base pieces by hand.
  • Add a square base, square heater, and square glass option to make sourcing and cutting easier. (The Reuleaux triangle option will still use a circle heater but will use circle glass instead of the specially cut basalt.)

I am sure I missed some things. Comment below and we can refine the new spec. Some day I will stop being lazy and mock up this design so we can talk and point instead of just talk.
Nicholas Seward
 
Posts: 738
Joined: Mon Nov 25, 2013 10:41 pm

Re: GUS v2 Goal Setting/Discussion

Postby Rasputin73 » Fri Aug 22, 2014 12:48 am

Well that's a lot of good stuff to chew on!
  • The Hub - I really like the idea of enlarging the hub. Being able to mount the feed motor direct is a good option. Providing room for axial fan cooling is great.
  • Lazy Susan -I'm a little concerned about the printed lazy susan idea. Even if you are using perfect balls, the tracks being printed may cause resistance, binding, and wobble at the hot end. We still haven't been able to print spools that are round enough, so that's really where my concern is coming from. Is the reason for printing it because bearings that big are expensive, to save weight, or just to make more parts printable?
    I know aluminum lazy susans are available on sites like alibaba for fairly cheep in a large variety of sizes. Maybe designing around their specs would be an option?
  • Probe - Providing a standardized mounting position for the bed probe and adding it to the vitamins will make it part of the standard. I know the software might not be completely ready for it, but it will be eventually and having the hardware in place in advance will make life easier

I'll keep adding as I think of stuff. :P
Rasputin73
 
Posts: 28
Joined: Wed Jan 01, 2014 11:43 pm

Re: GUS v2 Goal Setting/Discussion

Postby Nicholas Seward » Fri Aug 22, 2014 1:36 am

The FSRs will let the hot end act as the probe.

I will think on using stock lazy susans. I will also test printed ones. Being that big with the averaging effect of multiple balls should make printed ones pretty good.
Nicholas Seward
 
Posts: 738
Joined: Mon Nov 25, 2013 10:41 pm

Re: GUS v2 Goal Setting/Discussion

Postby sane » Mon Aug 25, 2014 2:34 am

I held off on building a GUS Simpson on the basis it wasn't accurate enough to print itself (which AFAIK is kinda key to Reprap). Recently I've looked back, to find little progress on that front - so I've been looking at alternatives (like Morgan).

To me, goals should encompass:

  • it should be dimensionally accurate enough to print itself, and anything else you want.
  • it should have a build volume bigger enough that it isn't a limitation (eg warping takes over)
  • it should be self-calibrating/setting up
  • it should be adaptable to different extruders, as well as multiple extruders
to this I add a few personal tweaks and interest areas:

  • it should be robust to inaccuracy in it's build, components, etc. This ties in with the above self calibration/setting up - stepping a motor 463 times is fine, but you do kind of need an independent measurement approach to know what that has meant in the real world and to close the loop. A big missing bit in most 3D printers I feel, and a way to make good with cheaper components.
  • should be possible to explore more cutting edge stuff, like dual extruders for supports, colour, and particularly for my interests, powder 3D printing. That tends to mean the ability to move the bed vertically.
  • should be sexy :mrgreen:
Some thoughts -

In response to your plan, can I ask why you look to a 100mm hub? What drives the size? Why not 200 or 250mm in diameter? How do the errors go with diameter? The bigger the hub in this Mk2, the more options that will fit.

At the same time, the bigger the hub, the more weight to lift and more inertia. What other mods are needed to deal with this without those lines slipping?

Finally, when I look at a GUS Simpson printing, standing back, it seems to have a lot of motion of the arms for a relatively small change in the head position. One of the advantages of something like the Morgan design is the arms magnify the movements of the motors in spacial position - potentially giving it a very large build area and better energy efficiency(?) GUS Simpson is very much an "outside in" design, compared to the Morgan's "inside out" design. Is there in inverse of the GUS's kinematics?
sane
 
Posts: 12
Joined: Tue Jan 14, 2014 9:47 pm

Re: GUS v2 Goal Setting/Discussion

Postby Nicholas Seward » Mon Aug 25, 2014 4:16 pm

sane wrote:I held off on building a GUS Simpson on the basis it wasn't accurate enough to print itself (which AFAIK is kinda key to Reprap). Recently I've looked back, to find little progress on that front - so I've been looking at alternatives (like Morgan).

Morgan is a very good machine and more importantly ready. GUS is still evolving.

I myself have been working on a 1-arm SCARA that builds upon the things that make the Morgan such a good printer.

GUS is aimed at solving different problems. GUS can be almost entirely printed, the bed is stationary, and there are no linear rails. I think this is key to scalability. I envision house printers that use this geometry. Most printers scale very poorly. Let's look at scaling costs

Cost=Constant*length^scale_factor

If you fit Cartesian printers to this formula and solve for scale_factor then you get a number between 2 and 3. When I look up what it would cost to scale a Simpson, the scale_factor is between 1 and 2. That is huge!!!. (To really drive my point home, do any heavy machines used in construction use much linear motion hardware? Not really. Most everything is built with simple pivots and wheel.)

sane wrote:To me, goals should encompass:
  • it should be dimensionally accurate enough to print itself, and anything else you want.
  • it should have a build volume bigger enough that it isn't a limitation (eg warping takes over)
  • it should be self-calibrating/setting up
  • it should be adaptable to different extruders, as well as multiple extruders
to this I add a few personal tweaks and interest areas:

  • it should be robust to inaccuracy in it's build, components, etc. This ties in with the above self calibration/setting up - stepping a motor 463 times is fine, but you do kind of need an independent measurement approach to know what that has meant in the real world and to close the loop. A big missing bit in most 3D printers I feel, and a way to make good with cheaper components.
  • should be possible to explore more cutting edge stuff, like dual extruders for supports, colour, and particularly for my interests, powder 3D printing. That tends to mean the ability to move the bed vertically.
  • should be sexy :mrgreen:

Dimensional accuracy is on the top of my list. Items 2,3,4,5,6,8,9, and 15 all will work to improve this.

The size is already great. You can print 200mm + long parts.

I also agree that it needs to be more flexible and support any hot end/extruder that someone wants to use. See item 11 and 12.

Powder printing doesn't really play to GUS's strength. That really needs to be a printer that has a dedicated XY stage.

Closing the loop would be fantastic. Implementing it will be a bit of a challenge. I am working in that direction. However, all the other cheap printers do just fine without closing the loop so I am working at this from both direction. 1) Close the loop cheaply. 2) Make the open loop more accurate.

sane wrote:Some thoughts -

In response to your plan, can I ask why you look to a 100mm hub? What drives the size? Why not 200 or 250mm in diameter? How do the errors go with diameter? The bigger the hub in this Mk2, the more options that will fit.

The main reason to keep the hub size down is to save plastic, make it lighter, make it easier to stay rigid, and make it more printable. 100mm would allow for a double extruder if someone wanted to do it. Bigger than that, you would probably want to switch to an inverted Simpson. (Have the bed move and keep the hot end static.)

The bigger the hub is compared to the shoulder the better behavior you get close to the shoulder pivots. However, it makes the math more demanding on the processor if the shoulder and hub offset aren't the same. I will look at expanding the shoulders but I think they will be too big if I match the offsets.

sane wrote:At the same time, the bigger the hub, the more weight to lift and more inertia. What other mods are needed to deal with this without those lines slipping?

Finally, when I look at a GUS Simpson printing, standing back, it seems to have a lot of motion of the arms for a relatively small change in the head position. One of the advantages of something like the Morgan design is the arms magnify the movements of the motors in spacial position - potentially giving it a very large build area and better energy efficiency(?) GUS Simpson is very much an "outside in" design, compared to the Morgan's "inside out" design. Is there in inverse of the GUS's kinematics?


Remember, the arms are very light. Visually it looks like a lot is moving around but the arms are mostly empty space. Additionally, experimentally we can generate high (>>standard printing speeds) speeds with relatively small steppers. Theoretically, speed and resolution is better distributed across the build volume than on the Morgan.
Nicholas Seward
 
Posts: 738
Joined: Mon Nov 25, 2013 10:41 pm

Re: GUS v2 Goal Setting/Discussion

Postby Rasputin73 » Mon Aug 25, 2014 8:00 pm

My thoughts on your suggestions...
sane wrote:
  • it should be dimensionally accurate enough to print itself, and anything else you want.
  • it should have a build volume bigger enough that it isn't a limitation (eg warping takes over)
  • it should be self-calibrating/setting up
  • it should be adaptable to different extruders, as well as multiple extruders

All of these are important in the long term, but dimensional accuracy and calibration are both mostly software/firmware issues and are inter-related. Granted, having hardware integrated to support the calibration is the first step. Nicholas is planning on using FSR's for this purpose in GUS V2. I'd also note that RepRaps in general aren't all "self-calibrating", mostly because of the diversity of hardware, software, and firmware used. I'll agree that they should be by default though. That's a major difference between hobbyist level commercial printers and the RepRap offerings and an area that RepRaps need to catch up on.

The build area doesn't seem to be an issue to me. Nicholas addressed this well, but the ratio of size of machine to build area on the GUS is really exceptional compared to almost any Cartesian printer and it scales well. You specifically referred to warping though, and this might be a sticking point. Two common ways to mitigate warping on large parts are 1)heated platform, and 2) heated enclosure. Many commercial printers use a heated enclosure, but hardly any RepRaps do. We've already got the platform in GUS, but an enclosure isn't reasonable with this geometry. I suppose you could put a box over the whole thing, but I wouldn't want to. I'm just an interested lurker and don't have much direct experience with this, but that's most of what I've read about the issue.

Different/multiple extruders will be simplified by the larger hub proposed in V2, but the software is really the key for multiples. Having multiple nozzles with GUS means they have to be offset from center. With the hub rotation, that would be another major factor for software to handle. Feeding multiple filaments to a single nozzle is another solution that has been tried by many, but it has its own issues with retraction, purging, etc...
sane wrote:
  • it should be robust to inaccuracy in it's build, components, etc. This ties in with the above self calibration/setting up - stepping a motor 463 times is fine, but you do kind of need an independent measurement approach to know what that has meant in the real world and to close the loop. A big missing bit in most 3D printers I feel, and a way to make good with cheaper components.

Closed loop would be great, and it would in many ways reduce the calibration needs. Homing positions, calibration, and end stops are all ways to mitigate the lack of a true 'closed loop' system. I think there are a few reasons most printers aren't closed loop... 1)parts cost increases 2)the sensors needed must be calibrated also 3)software or firmware complexity increases dramatically.

The only reasonable solutions I've seen are using servos rather then steppers (or at least an add on encoder) to close the loop on the drive mechanism only, or computer vision techniques to verify the actual position of the print head. One stepper w/ encoder and driver solution I've found from LoboCNC is here. It integrates the motor driver and encoder for mounting directly to the stepper. It's worth looking into, but it will add cost and vitamins. I doubt the computer vision option could be implemented on the current controllers, and it is hardly a robust or standardized solution yet. If anyone has a cheap, accurate and reliable way to close the loop, that would be a huge improvement for the whole 3D printing industry. If it can be solved in GUS, it would give GUS a major advantage in the RepRap field.
sane wrote:
  • should be possible to explore more cutting edge stuff, like dual extruders for supports, color, and particularly for my interests, powder 3D printing. That tends to mean the ability to move the bed vertically.

  • All fun goals, but V2 is mostly an iterative improvement over GUS V1. For the most part it needs refinements and incremental improvements as the focus. Some of those improvements will probably provide capacity for later feature expansion like multiple filaments, but probably not by default in V2. If we get lost in adding 'bonus' features right now then I suspect 'mission creep' will limit progress. My understanding of the goal is that V2 is following the 'Tick Tock' design model that Intel uses. Alternating between major process changes and then refinement. Nicholas can correct me if I'm wrong about that. :D

    Once the dimensions of the V2 hub are done and finalized, we can all look into designing multiple extruders and color heads for it. Within the RepRap community, lots of work has gone into two and three color systems, color mixing, filament splicing, etc... There hasn't been a real stand out 'winning' design yet, but a solid and reliable 'true color' system would also give GUS a major head start on other designs. I just don't think it can happen in the current GUS hub due to space limitations, and we don't have the final spec for V2 yet.
    sane wrote:
  • should be sexy :mrgreen:

  • Well, that's all up to personal preference I guess! I think GUS is already the sexiest little printer I've seen. The way it moves really inspires me. The unique and elegant aesthetics get my creative juices flowing! :lol:
    sane wrote:In response to your plan, can I ask why you look to a 100mm hub? What drives the size? Why not 200 or 250mm in diameter? How do the errors go with diameter? The bigger the hub in this Mk2, the more options that will fit.

    Going to 100mm (give or take) will give room for direct mounting of the filament drive rather than a bowden tube (if it's desired), possible fan cooling down the central shaft, a larger variety of hot ends, and possibly multiple extruders eventually. It just gives more real-estate at the 'business end' for options. The trade off is that a much wider hub would mean less build volume and more weight, so there is a real limit to how big it should be.

    So in a nutshell...
    • iterate, refine, improve.
    • Include only new features that fix problems with the current GUS,
    • OR features that prepare the design for future improvements.
    • Identify priorities within those guidelines.
    That's my (super long) 2cents. I hope everyone feels free to correct any mistaken or ignorant thoughts I've presented.
    Rasputin73
     
    Posts: 28
    Joined: Wed Jan 01, 2014 11:43 pm

    Re: GUS v2 Goal Setting/Discussion

    Postby Rasputin73 » Mon Aug 25, 2014 8:15 pm

    Quick followup on closed loops systems. Here's a recent one referenced on Hackaday called ServoStock that's interesting. It's physically a Kossel delta derivative, but the magic is in the custom controller and encoders. Probably too much divergence from standard electronics for use in GUS, but interesting ideas. See here and here.
    Rasputin73
     
    Posts: 28
    Joined: Wed Jan 01, 2014 11:43 pm

    Re: GUS v2 Goal Setting/Discussion

    Postby Nicholas Seward » Mon Aug 25, 2014 10:09 pm

    @Rasputin73: I just have one happy correction. As the hub size initially increases, the build volume grows. That is one reason I am very willing to increase the size of the hub.

    I also had a chance to try to source lazy susans in this size range. I was unsatisfied with everything that I found. I will print some tests and see how it feels to my hands.
    Nicholas Seward
     
    Posts: 738
    Joined: Mon Nov 25, 2013 10:41 pm

    Re: GUS v2 Goal Setting/Discussion

    Postby sane » Tue Aug 26, 2014 2:48 am

    Nicholas Seward wrote:
    sane wrote:I held off on building a GUS Simpson on the basis it wasn't accurate enough to print itself (which AFAIK is kinda key to Reprap). Recently I've looked back, to find little progress on that front - so I've been looking at alternatives (like Morgan).

    GUS is aimed at solving different problems. GUS can be almost entirely printed, the bed is stationary, and there are no linear rails. I think this is key to scalability. I envision house printers that use this geometry. ...(To really drive my point home, do any heavy machines used in construction use much linear motion hardware? Not really. Most everything is built with simple pivots and wheel.)

    Yep, and one of the reasons for my last paragraph:
    sane wrote:Finally, when I look at a GUS Simpson printing, standing back, it seems to have a lot of motion of the arms for a relatively small change in the head position. One of the advantages of something like the Morgan design is the arms magnify the movements of the motors in spacial position - potentially giving it a very large build area and better energy efficiency(?) GUS Simpson is very much an "outside in" design, compared to the Morgan's "inside out" design. Is there in inverse of the GUS's kinematics?

    As it stands, for GUS, the machine always has top be bigger than the build area ("outside in"), whereas when you look at Morgan the potential build area is bigger than the machine (at least the XY part). To me the way to get efficient scalability is that kind of kinematic arrangement. What would you think the cost scaling factor of Morgan would be? For that matter, what would a 3D Morgan look like ... maybe three GUS arms meshed with a pantagraph and building from the top?

    I confess I can't work it out myself, but it feels like there's something in that space.
    sane wrote:The size is already great. You can print 200mm + long parts.

    Height is more constrained and area is non-square, meaning I'd feel happier with a bit bigger maximum dimension to encompass a cubic 200mm build volume.

    Nicholas Seward wrote:Closing the loop would be fantastic. Implementing it will be a bit of a challenge. I am working in that direction. However, all the other cheap printers do just fine without closing the loop so I am working at this from both direction. 1) Close the loop cheaply. 2) Make the open loop more accurate.

    I think there is a cross over in the approaches between more expensive manufacture to increase accuracy and close loop to mean you can deal with inaccuracy. I just wonder where it is.
    Nicholas Seward wrote:Bigger than that, you would probably want to switch to an inverted Simpson. (Have the bed move and keep the hot end static.)

    I did consider that, but the bed wafting around suggests issues with the printed object toppling over as speed increases.
    Nicholas Seward wrote:Remember, the arms are very light. Visually it looks like a lot is moving around but the arms are mostly empty space. Additionally, experimentally we can generate high (>>standard printing speeds) speeds with relatively small steppers. Theoretically, speed and resolution is better distributed across the build volume than on the Morgan.

    Are they? I can see the resolution and speed being inversely related with radius in the Morgan, but with the GUS there are three arms to shift to arrive at any point (3D error, rather than 2D) and the outside in kinematics suggest the arms always have to move faster than the head (eg they end up being a limit).
    sane
     
    Posts: 12
    Joined: Tue Jan 14, 2014 9:47 pm

    Re: GUS v2 Goal Setting/Discussion

    Postby Rasputin73 » Tue Aug 26, 2014 3:08 am

    Oh good!

    Nicholas Seward wrote:@Rasputin73: I just have one happy correction. As the hub size initially increases, the build volume grows. That is one reason I am very willing to increase the size of the hub.

    Thrilled to be wrong on that one. Thanks for clarifying.

    Nicholas Seward wrote:I also had a chance to try to source lazy susans in this size range. I was unsatisfied with everything that I found. I will print some tests and see how it feels to my hands.

    I'm generally unhappy with the pressed steel lazy susans you get at a hardware store. Their tolerances are really low. I'm more hopeful about the machined aluminum ones coming out of china like this one here. Just picked a link for an example image. Lots of sizes available, and the price is decent. Maybe thats what you tried, not sure.
    Once you start getting to really high precision ones they tend to refer to them as 'Slew Bearings" or "Ring Bearings", but at that point they get crazy expensive.
    Rasputin73
     
    Posts: 28
    Joined: Wed Jan 01, 2014 11:43 pm

    Next

    Return to GUS Simpson

    Who is online

    Users browsing this forum: No registered users and 1 guest

    cron