If you're a real nerd about belt layouts,

hacks
experimentation
gplus
development
(Sanjay Mortimer) #1

If you’re a real nerd about belt layouts, this is an interesting new approach to the CoreXY/HBot approach.
http://blog.fabtotum.com/blog/2014/04/04/belts-and-stuff-experiences-to-share/

(Sven Eric Nielsen) #2

Even though it’s indeed very interesting, it’s not really new. It’s more than a year old.
I’m working on such a layout since a few month now. Unfortunately I’m far away from finished due to some family things. But as far as I can tell this is the most promising setup I’ve ever seen.
Directly started to convert that system from belts to fishing line and some other modifications.

The approach from this site gives you the cleanest setup I’ve ever seen.

(Ryan Carlyle) #3

I’ve been running this belt layout on my CoreXY printer since, eh, maybe last August. But to be honest, there’s really nothing wrong with crossing the belts.

(Brook Drumm) #4

I’ve built multiple bots with this setup. I’ve used all the belt layouts except one and even a few more not listed. I’ve even done ball screw+belt hybrids (on same axis) which get really weird.

Conclusion: The shorter the belt, the less stretch you have = less problems. Not to mention it’s cheaper. In our printers and cncs, I often use a single stretch of belt, sometimes I drag the motor on the carriage- yes, adding weight. I am of the opinion that slow is fine, since milling and printing are both pretty slow. No one should ever do a speed demo without actually printing or milling - AND showing the end result. Theoretical speeds and relying on math to predict speeds and loads are great, but testing (many many hours) in the real world is the real proof.

$0.02
Brook

(Jarek Szczepański) #5

that’s the layout which I used in my SmartCore variation! https://goo.gl/photos/AykoPH9NwYdVrDdd8

(Whosawhatsis) #6

@Brook_Drumm The issue isn’t speed as much as it is acceleration/jerk. Extruders have pressure hysteresis (and pressure advance algorithms are tricky to calibrate, relying on viscosity which is material-dependent and nonlinear), so 3D printers will always work best if they can achieve infinite acceleration so that nozzle pressure can remain constant around turns. Back in the Cupcake days, the firmware tried to do this, and while the machines had a lot of problems and had to run really slowly by today’s standards, they lacked some of the problems we have now because they could keep nozzle pressure constant. For the mechanisms to handle high acceleration/jerk without skipping, you want to minimize the moving mass, and carrying motors around is just about the last thing you would want to do.

(Wolfmanjm) #7

yea hardly new at all as mentioned. I published this stacked belt approach back in march 2013, and I almost certainly was not the first. http://wolfmanjm.github.io/wolfhbot/

(Ryan Carlyle) #8

@Whosa_whatsis Agreed totally. The biggest benefit to CoreXY is stationary motors. But, additionally, the deflection caused by the stepper torque/error relationship is out of alignment with the deflection caused by rod flexing. So CoreXY will actually outperform a parallel Cartesian design (Slid3r/Dualwire-Gantry or Ultimaker style machines) because the acceleration/jerk deflections caused by the motors and rods are partially out of alignment and thus the resultant total flexure/ringing magnitude is smaller.

I still don’t think I’d use CoreXY for a serious CNC mill – it’s a hell of a lot of belt length to worry about stretching.

(James Zatopa) #9

Sanjay, this reminds me a bit about what we were discussing at the MRRF. Technically they do cross the belts, the main benefit looks to be that the belts are kept square. Another option is to do what the Fusion3 printer does and use a filament drive.

(Brook Drumm) #10

Acceleration can minimize jerk. Our pro drags s lot of motors. So does our Cnc. Henner Zeller has his screaming fast using his home-rolled beagle one black shield.

Tinyg is gonna add acceleration to the extruder flow. They expect zero retraction needed. I guess it will eliminate the need for constant pressure in the hotend.

(Ryan Carlyle) #11

@Brook_Drumm What do you mean by adding acceleration to the extruder flow? Are they talking about a pressure advance function like Sailfish uses? I really don’t know much about TinyG, so forgive me if I’m misunderstanding you.

The extruder rate for all the “standard” 3DP firmwares is typically slaved to the motion axes, so when the motion axes see acceleration slowdowns at corners, the extruder is given the same proportional acceleration slowdown to (theoretically) match pace. (Pressure advance functions make up for the non-linearities, but most firmwares don’t have effective functions at the moment.) I don’t think anybody’s really used constant-pressure extruders since the RPM extrusion days. It’s all volumetric 5D with the extruder its own CNC axis now.

(Brook Drumm) #12

You are right, but the bar Tinyg is trying to reach is high. Not just slowing the extrusion, but treating corners with special care as to not pile up. There is also a possibility to treat outer surfaces differently and even handle infill differently. They are in the planning stages. True calculation of volumetric flow is exciting but geeky. They are bringing better processing power and much more mature firmware to the party.

(Ryan Carlyle) #13

@Brook_Drumm is there anywhere I could read about what they’re doing? Calculating true flow in a constant-jerk motion system is much more complex than almost anybody realizes – the simplest possible volume correction is probably a third-order empirical fudge. (Sailfish JKN Advance uses a second-order empirical fudge for constant-acceleration motion). But I’d be happy to be wrong about that if the TinyG guys are onto something.

And if the attempts of the MachineKit guys to implement Velocity Extrusion are any indication, it’s extremely easy to violate Stratasys patents without realizing it.

(Brook Drumm) #14

These guys are pjs level mathematicians going deep deep deep down the rabbit hole. I can’t even tell you! Mostly because I can barely hang on when they explain in layman terms what they are planning. They are knocking on some significant IP but want the open source community to benefit. Patents aren’t my wheel house but as Printrbot gets into software, we will have to figure out some sort of defensive strategy to protect ourselves. Tinyg / “G2” is a project they developed for them- a separate company, but they have several approaches for me to work with them and compensate them for their work. The project is on github but it may be a few months until the new 3d printer stuff is public. It’s been a very long road and lots to do. I like the idea of hitching my wagon to extremely serious coders and leaders in the space of motion control.