Welcome! Log In Create A New Profile

Advanced

High speed delta's are often not that speedy at all

Posted by evandene 
High speed delta's are often not that speedy at all
June 02, 2016 01:45PM
True 100, 200, even 300 mm/sec are indeed possible but only with the right means.
The most important variable in the whole speed story is mass and acceleration.
If a mass (Arms, Effector, sliders, nozzles, fans, etc, etc) is high and acceleration settings are low, often the requested speed (defined in the slicer program) will never exceed by the simple fact that the print point to point tracks are to short.
Increasing the acceleration there is a high risk that the machine will stall one or more joints.

So having a high speed set in your slicer doesn't say anything about the real speed over the print surface.
My Delta with a BeagleBone Black and a BeBoPr++ controller and Machinekit software has a max acceleration of 800 mm/sec². When set to 1000 mm/sec² acceleration the joints will stall before max speed of 120 mm/sec.

Often for higher accelerations you need more torque by using a spindle driven system (pitch 8mm for example) or Nema23 steppers, or Nema17 steppers with gearbox (efficiency > 85%)

smileys with beer

Edited 1 time(s). Last edit at 06/02/2016 01:45PM by evandene.
Re: High speed delta's are often not that speedy at all
June 02, 2016 03:32PM
Nema17's no gear reduction, easily hits 2000, works ok at 4000. No stalling. I also run a direct drive extruder. Not sure what your issue is, but your conclusions are not universal.
Re: High speed delta's are often not that speedy at all
June 02, 2016 05:14PM
I use accelerations of 7200 mm/sec² with Nema17 steppers. It is be possible to go to at least 17000 mm/sec² with Nema17 (if you push them to the limits and your platform is not heavy).
Re: High speed delta's are often not that speedy at all
June 02, 2016 05:59PM
I use 300mm/sec travel speed and 3000mm/sec acceleration. A simple calculation shows that full speed is reached from rest after 15mm. So it does reach 300mm/sec on travel moves longer than 300mm.

Using 0.9deg Nema 17 motors with around 0.5Nm torque at 1.66A rated current, but actually running at only 1A, and Duet electronics with 24V power. The effector is light weight with just an E3Dv6 hot end, 40mm cooling fan and IR height sensor.

Edited 2 time(s). Last edit at 06/02/2016 06:05PM by dc42.



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

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].
Re: High speed delta's are often not that speedy at all
June 02, 2016 06:03PM
If you can't even get 1000mm/s/s accelerate then it's a poor build for one or more reasons.... It's not that it is a Delta.

Admit that I'm using Nema23's but I'm already at 6000mm/s/s with zero signs of ringing.
Need to have some belt joiner tensioner pieces made from Aluminium before going too much higher.
At 100mm/s print speed I'm it full speed in 1mm.

ANY printer with poor part selection will not be as speedy as thought or advertised
Re: High speed delta's are often not that speedy at all
June 02, 2016 07:11PM
Oh dear, this has a horrible ring of familiarity about it,

Basically what you have 'discovered' here is if your not using the maximum speed and power of a motor then you can optimise the system by changing the gearing ratio, this will sacrifice the maximum RPM of the motor

MAX power=torque x MAX RPM

MAX acceleration= torque/inertia

MAX RPM=(step frequency/steps per mm *60) /mm per rev



looking at the basic laws of motion

v=u+at tells us that a 3000mm/s^2 acceleration will reach 300mm/s in 0.1 sec

d=ut + 1/2 . at^2 says that in 0.1s a 3000mm/s^2 acceleration will travel 15mm

To accelerate from o to max and back to 0 the minimum move needs to be 30mm


BUT, we are talking about printing speed not max travel speed I currently print at 100mm/s with 1200mm/s^2 acceelration so...

max speed in 0083s

min travel=8.33mm

If I play with the accelerations and use 3000 instead of 1200 (my setup is very similar to DC42s then.....

max speed in 0.033s
min travel = 3.33mm


However, lets not forget that just talking about axis speed is a waste of time, in reality what we need to talk about is effector speed, and whilst in theory increasing axis speed and acceleration is reasonably straight forward, what happens when you accelerate the effector is you impart bending moments in the arms, which in turn will impart a frequency oscillation back into the effector, ultimately resulting in reduced printing quality, which is ironic really because you started this chain of thread on a way to increase printer accuracy to sub perceptible levels by sacrificing speed, now you are talking about increasing speed by sacrificing print quality....somewhere you would expect there to be a happy medium.... I wonder how far off it the current designs are?

but ultimately you are still talking about building a screw based printer and trying to post rationalise it.

Edited 3 time(s). Last edit at 06/03/2016 07:57PM by bgkdavis.



RepRapPro Mendel 3 Tricolour
RepRapPro Fisher
-Carbon Arms
-Easy adjust Carriage+effector
-axis stiffness mods
HE3D -600 delta
-Duet 0.8.5
-PanelDue
-DC42 Height probe
-RobotDigg metal components
Simplyfy3D
RS Design Spark CAD
Re: High speed delta's are often not that speedy at all
June 03, 2016 03:48AM
Nice approach, and helpful, thanks.

For me it's obvious to talk about max accelerations and max speed of the effector.
And of course product quality; overall dimension tolerance of 0.1mm tolerance FIELD is a must for me. (Cpk 1.67)

I admit that although my own design and build is very precise, with very low inertia of effector and arms, I need to improve acceleration capabilities.
Part of my problem comes from my Glacier Teflon Bearing Bushes used in combination with my D=16 pillars bearing blocks.
Misalignments of my pillars of a few hundreds of a mm could have big influence on friction.
The easiest way is changing my design and use NEMA23 motors with decent torque.

Edited 1 time(s). Last edit at 06/03/2016 03:49AM by evandene.
Re: High speed delta's are often not that speedy at all
June 03, 2016 04:10AM
4000 mm/sec. sec acceleration for a NEMA17 with a Pull-out Torque of 0.42NM motor is simple not possible for a belt driven system with carriages, arms, effector and friction.
Typical "bigger is better talk".

Pull-out torque
The stepper motor pull-out torque is measured by accelerating the motor to the desired speed and then increasing the torque loading until the motor stalls or misses steps. This measurement is taken across a wide range of speeds and the results are used to generate the stepper motor's dynamic performance curve. As noted below this curve is affected by drive voltage, drive current and current switching techniques. A designer may include a safety factor between the rated torque and the estimated full load torque required for the application
.
Re: High speed delta's are often not that speedy at all
June 03, 2016 06:06AM
Quote
evandene
4000 mm/sec. sec acceleration for a NEMA17 with a Pull-out Torque of 0.42NM motor is simple not possible for a belt driven system with carriages, arms, effector and friction.
Typical "bigger is better talk".

Pull-out torque
The stepper motor pull-out torque is measured by accelerating the motor to the desired speed and then increasing the torque loading until the motor stalls or misses steps. This measurement is taken across a wide range of speeds and the results are used to generate the stepper motor's dynamic performance curve. As noted below this curve is affected by drive voltage, drive current and current switching techniques. A designer may include a safety factor between the rated torque and the estimated full load torque required for the application
.
If that's what you need to tell yourself to rationalize why your system doesn't perform well that is fine. Perhaps you might try backing that up with some math, because the system I have in front of me works fine at that accelleration.
Re: High speed delta's are often not that speedy at all
June 03, 2016 11:39AM
Quote
evandene
4000 mm/sec. sec acceleration for a NEMA17 with a Pull-out Torque of 0.42NM motor is simple not possible for a belt driven system with carriages, arms, effector and friction.

My belt driven printer proves otherwise since I use 7200 mm/s² accelerations and 120 mm/s printing speed (250 mm/s travel speed). You need to revise your calculations.

If you are the guy who wants to build a lead screw delta then I already told you you should try to design your printer for at least 150 mm/s print speed and at least 10000 mm/s² accelerations. This is needed if you want your printer to be quick. I especially highlighted the acceleration problem with lead screws as a challenge you will face. If you cannot achieve high speeds/accelerations then you should rather aim for a Cartesian printer. If you are somebody else then disregard this paragraph.
PRZ
Re: High speed delta's are often not that speedy at all
June 04, 2016 04:22AM
On a Delta, accelerations which are discussed here are effector acceleration, meaning that at maximum diameter, the carriage acceleration will be 2.5 to 3 times higher due to geometry. That is a major constraint of delta design, which explain why designing a delta is even more than for other printers type, chasing weight of mobile parts. That is why wheels shall be preferred to sliding ball bearings for weights reasons, why effector shall be the smallest possible, hotend shall be the lightest and you shall not have an extruder on the effector.

The dynamic load on mobile parts are linearly linked to weight and accelerations, so increasing the weight have the same effect as increasing the acceleration.
To make comparisons, gravity acceleration is 9 810 mm/s^2.

Why do we need high acceleration on a Delta ? Most delta are using Bowden tubes, meaning the filament is compressed of a significant amount in the tube, so still pushing material even after extruder slowdown, stop or retract. The rate of extrusion decrease with the pressure, but this is far from being instant. This is why we have blobs on angles of the parts and artifacts around transversal holes (X or Y). The longer the Bowden, the worse. Flying extruders are a good compromise, maintaining low inertia (the extruder get much lower accelerations than the effector, so implying low dynamic loads), with a lot less filament compression.

A working pressure advance control may help, especially for long Bowden tubes and low acceleration machines, but this is delicate to tune, the values depend from the type of filament and this seems difficult to implement.

With high acceleration, there is high loads so more stress on the machine, more vibrations, more wear and more noise. With carriage acceleration exceeding the gravity, any backlash will create an undesirable jump and will quickly wear parts.
High acceleration imply stiff parts, but if the stiffness is obtained while increasing the mass, the overall result may be negative.
This is why design is critical. It is very difficult to design a machine with plastic parts having a stiffness to weight ratio better than aluminium because aluminium stiffness is 10 to 20 time the plastic, for less than three time the weight. But heavier parts add stress, so flexibility to other components and notably belts. Finding a good compromise is a difficult exercise.

There is two kind of accelerations:
- The sustained acceleration, as defined in your parameters
- The acceleration when changing speed. There is a parameter which allow the machine to start at a given speed or change speed without ramping up, sometimes called jerk. That makes the theoretical acceleration infinite. However, the machine and its transmission and the steppers themselves do have flexibility and the real acceleration depends of the machine construction. It is important to note that for a well built machine, this jerk parameter shall be lower than for a weak machine. The stepper position itself is not perfectly rigid, especially when microstepping. Using a 0.9° stepper instead of a 0.8° double the stepper stiffness.

The small Fisher Delta , as tuned from manufacturer, have 4000 mm/^2 acceleration. It is true that it is not quite sustainable at the maximum specified diameter, but for miscellaneous reasons, the maximum diameter is not quite usable on this printer. The Fisher is rod based machine (rod diam 8mm), with undersized plastic carriage, cantilevered balls attachs on effector and carriage and an effector weighting a bit more than 100g, heavy compared to a stock Kossel. Also, the rod slide bearings are heavy compared to wheels, but overall, the carriage weight is reasonable. And the arms are in acrylic, heavy for this machine size. It print at 80mm/s, travel at 200mm/s with steppers on 0.22 N.m under 19V. A remarkably optimised motorisation.
This is a quite flexible machine, designed for low cost, which however can sustain reasonable accelerations. One of the reason is this is a small machine, which does have significant advantages for dynamic behaviour.

I have another printer of own design, the D-Box, This is a machine with plastic carriages and plastic stepper supports in PETG, so more flexible than PLA. Balls on effectors are in printed cups, so relatively stiff. Arms are quite stiff longitudinally. My belts are average quality. My effector weight 200g, which is on the heavy side due to the installation of a kinematic positioning system. Carriages are on ball bearings, very stiff, and rails are quite stiff in bending. The structure is a wood box, extremely stiff. Overall, a machine with good characteristics, but you may do better with a machine having carriages and steppers supports in aluminium. From what I see, the largest flexibility came from the belt, carriage ball cup supports and steppers support. It can print at 150mm/s and travel at 250mm/s under 24V with steppers of 0.42 N.m and 0.9°. This means that same machine with steppers 1.8° could approximately do 500 mm/sec.

This machine is yet tuned with acceleration of 6000 mm^2, while I have tested it at 8000. Max speed ratio at largest diameter is 2.5, so is the maximum carriage acceleration, hence 6000x2.5 = 15000 mm^2.

I have modified my arms, which were aluminium with balls screwed in aluminium on both end. The new arms are combined carbon fiber and aluminium, but the ball screws are glued on carbon rod and aluminium tube, so there is no longer any plastic part involved in the longitudinal flexibility. This has significantly improved the longitudinal stiffness (and reduced the weight) and while installing these arms, I observed a stupefying phenomenon, my effector was jumping in its support at each layer change. The mobile part of the effector weight 120g and the retaining force is 1kg. Do the math, to have it jumping, it needs 1kg/0.12kg x 9810 = 81 750 mm/s^2. That is incredible for a belt based system. I assume there was some dynamic phenomenon but that is anyway really huge. Reducing the jerk value eliminated the problem.

If a machine as weak as the Fisher can sustain an acceleration of 4000 mm/s^2, There is no reasons why a well built machine with twice the stepper torque cannot use two or three time this value, provided mobile parts remains light.

A well built machine, probably one of the most stiff which can be done in plastic is the Griffin. On Forum, his builder tells that he have experienced acceleration of more than 10 000 mm/sec^2. With an effector much lighter than the one I have on the D-Box, this is a credible value, especially with the smaller version of this machine.

Your belts are more than capable to support your parts, so sustain the gravity acceleration of 9810 mm/s^2, why shouldn’t they be capable to give three time this value, especially with light mobile mass ?

So, yes, you could have high acceleration on a delta with belt transmission. However, the belt is one of the weak point, and good quality or larger belts may improve the system.
A important question is the possibility to use polyurethane belts with steel wire reinforcement (the white one). The problem is that steel reinforced belts need larger pulleys, and so, larger torque on the stepper, so that will drive to more construction difficulties. Also, larger pulley is equivalent to reduce the stepper stiffness and increase required torque. That may be possible with stepper driver capable to really supply 2A or more. The precision and stepper stiffness could be maintained to reasonable values if you are using 0.9° steppers.

Using a value of 4000 mm/s^2 will get carriage acceleration below gravity for most of the printable diameter, which will help avoiding backlash troubles with parts having play.

Edited 1 time(s). Last edit at 06/04/2016 04:25AM by PRZ.


Pierre

- Safety [reprap.org]
- Embedded help system for Duet and RepRap Firmware [forums.reprap.org]
- Enclosed delta printers Lily [rouzeau.net] and Lily Big [rouzeau.net]
- OpenScad delta printer simulator [github.com]
- 3D printing on my site [www.rouzeau.net]
Re: High speed delta's are often not that speedy at all
June 04, 2016 06:31AM
Thanks Pierre,

I have all weights, and inertia data available and will walk through my design step by step. My Delta arms of 296 mm center to center are of Aluminum D=12x1 and the ball joints at the end are D=14 mm, ... so quite heavy and possible an area for improvement. Furthermore My stepper motors are 12V 1.7 A, 0.35NM, 1.8º and these should most probably also get upgraded. The effector is stiff and a light weight design with only an E3D system mounted.

Thanks again
Re: High speed delta's are often not that speedy at all
June 04, 2016 09:22AM
@PRZ

You sort of hinted at the fact that the entire mass of the system driven by the steppers needs to be taken into account...... I'm going to say you must take all the mass into account.

I also using the term mass as you can counterweight things like flying extruders but they still have mass to speed up and stop.

I know my effector set up weighs in around 1/2 to 1/3 of the weight of any single carriage/rod pair.
I did weigh the fully kitted effector but cannot remember what it was now.

@evandene

Heavy is not always bad, but it has to all work together and as always heavy is relative to the size of the machine.

Very little in my Delta would be considered lightweight but everything is SOLID, the entire frame is now braced with external panels that are actually there for the heated enclosure set up but have served to brace the framing enabling speeds to go up.

I do not disagree with the mantra of lighten the effector.
Re: High speed delta's are often not that speedy at all
June 04, 2016 09:39AM
First and big mistake found in my lineardelta.ini file.
As I mentioned at the start of this topic I'm using a Beaglebone Black hooked up with a BeBoPr++ controller and DRV8825 drivers.
The software is a version of LinuxCNC "machinekit".
In the lineardelta.ini file and related PRU file, max step frequency is defined. My MAX STEPGEN VEL and MAX STEP ACC was set wrongly.
I'm not yet done with my corrective actions but my linear acceleration is already at 2000 mm/sec^2 and 100mm/sec print speed for the perimeters. Printed parts are within tolerance of 0.1mm.
Will keep you informed about all corrective actions.
Re: High speed delta's are often not that speedy at all
June 04, 2016 10:48AM
@PRZ: Interesting information. And after reading it I'm glad again I never tried magnetic joints :-D

I can contribute a bit about smooth rod based deltas. Stuff I learned by heavily modifying the "classical rostock".
One needs to remove all the noticeable backslash. This typically means higher quality diagonal rods (in my case carbon fibre with Ø 7mm PM-Jet ball joints), good selection of linear bearings and better idler, carriage, and platform design.
The rest is about stiffening the printer frame and rods/belts (1.8°, 50 N/cm steppers, 12.7 mm pulley diameter):
  • if you have Gates GT2 glass core belts then their maximum error is at most 4.2 mm
  • Ø 8 mm smooth rods contribute with at most 2.3 mm to the dynamic error (this is only pure dynamic error from accelerations, not the twisting errors you get on the rods because your hotend is dragging over bed and twisting the platform/smoothRods)
  • steel core T2.5 belt (this one has 4 * 7 steel filaments, each having 0.05 mm radius) - maximum error is 1.7 mm
  • maximum stepper position error is 0.1 mm
These are the very worst cases which rarely happen since we do not run the printers at the very edge of their capabilities (just before they skip a step). But it is useful to give you idea how big an error of the given part is compared to the rest.
Re: High speed delta's are often not that speedy at all
June 04, 2016 11:50AM
Second issue found in my lineardelta.ini file found.

In the Machinekit lineardelta.ini file In the section TRAJ (trajectory) of this .ini file you set your FERROR or Joint Following Error. A little strange when using steppers but if you like to read more about it please follow this link. Following Error definition
It looks as if I did gave FERROR a to low value in this .ini file because after the corrective action taken and after hour printing now there was no "joint following error" anymore. Acceleration was set at 2000mm sec.sec Perimeter Velocity was set at 80 mm/sec and travel at 150mm/sec. FERROR value from 2 to 6.
Next step will be 3000 mm/sec.sec and a Perimeter print speed of 100mm/sec
will keep you posted

Edited 1 time(s). Last edit at 06/04/2016 11:51AM by evandene.
Re: High speed delta's are often not that speedy at all
June 05, 2016 04:52AM
Thanks everybody for the input, very helpful and you all helped me to move my ignorance boundaries a little.
I learned that first of all, I did needed to know a little more about the insides of LinuxCNC (Machinekit) the software I'm running. That the guys who did build the out of the box working software did a good job is obvious but if one likes to improve the machine performance you need to understand a little about LinuxCNC; I took shortcuts there and I shouldn't have done that.
I optimized FERROR settings (Joint Following error value) from there I could start maximizing linear speed settings and acceleration settings. 4000 mm/sec² is what did work for me. I can go higher but than I need to optimize first the inertia of the mechanics in motion. the max 4000 mm/sec² and a perimeter max speed of 100 mm/sec works fine for my Delta, keeping parts nicely within 0.1mm tolerance.
Sorry for my ignorance at the start of this topic,
Thanks again all of you

Edited 2 time(s). Last edit at 06/05/2016 05:40AM by evandene.
Re: High speed delta's are often not that speedy at all
June 05, 2016 05:21AM
great to hear evandene

Cheers
Re: High speed delta's are often not that speedy at all
April 30, 2017 03:41AM
I decided to upgrade my DELTA printer from 1.8angle to 0.9 angle NEMA 17 steppers what should contribute to more smooth and precise printing.
I'm running Machinekit on a Beaglebone Black and a BeBoPr++ controller.
After installing the new stepper motors my delta.ini file needed to get updated too of course but I'm struggling again with acceleration settings.
I needed to lower my acceleration and max speed to half; speed 200 to 100 and max acceleration from 3000 mm/sec2 to 1000 mm/sec2
Does anyone has a suggestion how to setup for these new steppers?
Thanks
Re: High speed delta's are often not that speedy at all
April 30, 2017 05:16AM
I do not use Beaglebone and BePoPr++ so I cannot help you with the configuration.

But if the only thing you changed is the stepper motors from 200 to 400 steps per rotation then you definitely can keep the original acceleration and possibly also the original maximum speed.
The point is that:
  • Maximum acceleration is not influenced by step rate. It is influenced by weights of the moving parts and the maximum stepper torque. If you changed only the steppers then you need fiddle with maximum acceleration only if the new steppers have smaller torque or bigger moment of inertia (heavier rotor). I.e. also something else than single-step angle changed.
  • Maximum speed may be influenced if your firmware is not able to provide the new maximum step frequency (not probable with Beaglebone) or if your voltage is not high enough to support the new maximum step rate. You may need to increase the stepper voltage to keep high speed if you did not have enough voltage reserve or if the new steppers do not have significantly smaller inductance.
Re: High speed delta's are often not that speedy at all
April 30, 2017 08:54AM
To get good speeds with 0.9deg motors you need to chose motors with a reasonably low inductance and use at least 24V power. See [duet3d.com].



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

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].
Sorry, only registered users may post in this forum.

Click here to login