War against ripples

@Mechabits ok cool yea I thought not, just sounded odd. Thanks for clarifying. Dont mistake real for fake and vice versa, easily done.

---

Flex3Drive.com

unless of course your watching a politicians lips move, then you know it's all fake.

Quote
DjDemonD
Okay maybe I'm being slightly pessimistic but Marlin is frequently quoted as only being able to generate 40khz step pulses from a mega/ramps board. So you're using repetier? I've little experience with it, but to do the moves you quoted that's 75khz step pulses, so that seems reasonable from what I'm told it's code is more efficient.

But on a delta is a pure z move really a good test? Whilst moving up/down is the worst case in terms of maximum step pulses to be generated, the processing requirements to calculate the movement are very little, all motors move by the same amount to achieve pure z movement. Any other movements require a lot of square root calculations which load the processor . How fast can that same machine (on a dry run) do a circular spiral vase, this is a more useful test, given that you almost never do a large pure z move when printing?

I've been away from home for a week, and just got a chance to do that test. I made a cylinder with a perimeter of 100 mm, and printed it in spiral vase mode. The problem I ended up having was that the SD card couldn't be read fast enough to keep the buffers full, and I ended up having to reduce the number of polygons in the slice to get the speed to stay at a consistent 100 mm/s. This wasn't a good test of top speed of the motors.

So I sliced a 25mm cube (with a perimeter of 100 mm) and did a dry run (normal print, just without filament in the hotend). With acceleration at 6000 it seemed to be able to travel at pretty close to 200 mm/s. I took some videos and will upload them if there's interest.

The buffer issue is something I'd never seen before, but then again I don't actually print that fast so it's not a problem in normal prints. I was a bit surprised that my delta running RAMPS could get to 200 mm/s, I was only expecting 150 or so, but I guess in purely horizontal movement one tower won't be going very fast.

That's interesting share the stl I'll see if I can max out the duetwifi.

Edit - LOL just read this properly - 25mm cube think I can manage that.

Edited 1 time(s). Last edit at 03/25/2017 05:49PM by DjDemonD.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

Well I have got it moving at 330mm/s but I am not sure I dare to go faster this is on 1/16 to 1/256 interpolated microstepping.

I can try other modes. There's more mass now I've set up a dual flying extruder.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

Quote
DjDemonD
Well I have got it moving at 330mm/s but I am not sure I dare to go faster this is on 1/16 to 1/256 interpolated microstepping.

I can try other modes. There's more mass now I've set up a dual flying extruder.

I currently run my delta at x64 microstepping. The steps/mm is 800.

---

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].

Thinking about trying a few other modes, I did have an issue with probing using the piezo probe at anything above 1/32 it doesn't trigger properly. I found this out after you suggested on Duet forum to try higher modes when probing to increase accuracy. So I probe at 1/32 and revert to 1/16th interpolated to print (just sounds better).

That being said how fast can you get your machine moving at that microstepping rate? Its not a silly contest but if Nebbian can get his 8bit controller machine up to 200 mm/s I'm keen to see how much quicker Duetwifi actually is. Of course printing at these speeds exceeds the maximum volmetric rate of filament melt on any hotend I have but its interesting to know.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

When Michael Hackney did a lot of work on accelerometer-based bed probing, he found that using 256x microstepping worked best (better than 16x interpolated to 256x). This was because the smoother motion allowed him to use a higher sensitivity, so the nozzle didn't have to hit the bed so hard, and soft bed materials such as PEI gave a more consistent trigger.

I'll check the maximum movement speed when I am back at the printer. AFAIR I chose 64x because 128x didn't give me 200mm/sec Z movement.

Does nebbian's machine use 0.9deg motors?

---

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].

Is this still a war against ripples?

Edited 1 time(s). Last edit at 03/26/2017 04:46PM by Hatmpatn.

Kinda in that we worked out what you need to get rid of them was a 32bit controller.

Just trying to work out now how much more of a benefit (or not) one is.

Edited 1 time(s). Last edit at 03/26/2017 06:42PM by DjDemonD.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

Quote
dc42

Does nebbian's machine use 0.9deg motors?

Yes.

0.9 degree motors, Ramps, Repetier, 1/16 microstepping, 160 steps/mm.

I know that 32 bit boards are faster, there's no contest there. I did the test because many people keep repeating the myth that 8 bit boards can't move a delta fast. I think that 200 mm/s is fast enough.

I agree, 200mm/sec is fast enough for a Delta printer of normal size.

---

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].

What would be really interesting is to see what the thread creator icefire experiences after upgrading to a 32-bit controller board.

Edited 1 time(s). Last edit at 03/31/2017 05:38AM by Hatmpatn.

Well I am very curious as well but haven't tried it yet. This seems to be the most expensive experiment after building a CoreXY from scratch.

What I know so far is that there are almost no visible ripples at 0,2mm layer height. So If I want to go lower maybe it would be a good idea to use a smaller nozzle. I guess that there is a correlation between the nozzle size, the layer height and the ripples.

I also tried switching to repetier. Apart from the fact that I needed a few hours to figure out how the auto bed leveling should be set up there is no difference in print quality.

So it would be really cool to try a 32 bit board but there is also the cost factor and, which is by far the worst, there is no guarantee that it is going to get better.

---

Self-sourced Mendelmax 2.0-based Reprap Machine -- Ramps 1.4 & Mega 2560 -- DRV8825 (Z@1A, [email protected], [email protected], E@1A) -- genuine E3D v6 direct setup -- 350W custom silicone heated bed -- ABS 1,75mm -- Marlin 1.1.0-RC7 -- Cura 15.04.6

If it makes any difference I was very reluctant to use 32bit mainly because I was unsure if it would be good value. Now I have a duetwifi and a re-arm and they are worth every penny.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

I am not a fan of Chinese knock offs, but maybe it could be an alternative to use an Arduino Due clone (15 USD) and a Ramps-FD (20 USD) in order to just see how print quality is changed by going 32-bit?

---

Self-sourced Mendelmax 2.0-based Reprap Machine -- Ramps 1.4 & Mega 2560 -- DRV8825 (Z@1A, [email protected], [email protected], E@1A) -- genuine E3D v6 direct setup -- 350W custom silicone heated bed -- ABS 1,75mm -- Marlin 1.1.0-RC7 -- Cura 15.04.6

I've heard nothing but bad things about ramps FD. Probably better off with MKS-Sbase (I know, I know) but money wise its the cheapest route to smoothieware unless you get a re-arm (and already have a ramps/drivers).

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

If you want to use the DUE use the radds shield (63$ makerfarm in the US) Not the rampsFD, the old version got clone and QC is bad, the new version was suppose to come up to market but never did.

+ with radds you can use marlin/repetier and RRF (same as duet). So you can compare 16 bit with 32 bit using the same firmware or go RRF and see the higher quality when using a proper 32bit firmware with better segment calculation VS a modding of 8 to 32 bit firmware.

there is no point. For this kind of money + shipping from the US to Europe I'd get a duet.
A cheap alternative is mks sbase from aliexpress but... I don't know if it works well.... and it uses the controversial DRV8825..

---

Self-sourced Mendelmax 2.0-based Reprap Machine -- Ramps 1.4 & Mega 2560 -- DRV8825 (Z@1A, [email protected], [email protected], E@1A) -- genuine E3D v6 direct setup -- 350W custom silicone heated bed -- ABS 1,75mm -- Marlin 1.1.0-RC7 -- Cura 15.04.6

If you want a Duet you can surely get one in Europe since they're made here and the shipping will be even less.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

Well if you are in europe there better place and price to get the radds since it made in germany, But whatever you do avoid the drv8825 at all cost they have a very bad low current zone, result in missing step in slow and small movement, those movement we do them a lot on delta when doing a straight line one carrier is moving slowly and only a couple MM so we go right into that low current zone.

for driver , Stick with 4988 or go raps 128 (thb6128) or SD6128 or the tmc 26 something of the duet wifi.

finally there only 2 firmware that work well in 32 its rrf and smoothie , my opinion (rrf user) is RRF is very good and always add new stuff to firmware (for Duet board and even for radds board) and its always keep up to date, firmware are release almost every week or two with new stuff, you cant go wrong with that other than that they similar I guess. The other thing to check out is the hardware option: onboard driver/socket the number of them and stuff like that.

also beware that mks board often dont have full support from smoothie dev but I guess you can count on the mks user to help, without entering that dangerous slippery subject some statement was made about this I guess a little search around will tell you more detail about that.

All in all, I guess that printers with moving beds will always have the problem to some extent.

I have another idea: How about using a leadscrew for the Y axis instead of the GT2 belt? My heated bed is 320mm long in Y and the physical length of the axis is 420mm. In addition to that the Y carriage is pretty heavy so maybe the belt is just too elastic for this kind of weight. Maybe a leadscrew will eliminate the issue as it has no elasticity.

Any thoughts on this?

---

Self-sourced Mendelmax 2.0-based Reprap Machine -- Ramps 1.4 & Mega 2560 -- DRV8825 (Z@1A, [email protected], [email protected], E@1A) -- genuine E3D v6 direct setup -- 350W custom silicone heated bed -- ABS 1,75mm -- Marlin 1.1.0-RC7 -- Cura 15.04.6

You can but I always presumed that the issue with leadscrews for x and y (or all three axes on a delta) was that they are slow unless you go for something with a low number of threads per unit length, which require quite a bit of torque to drive and also then have lower resolution. Maybe with very powerful 0.9 deg motors you might get reasonable speed and resolution.

It works for cnc as you want high thrust not high speed.

Edited 1 time(s). Last edit at 07/02/2017 09:15AM by DjDemonD.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

Well I could select a suitable pitch. Alternatively I could try a wider GT2 belt (15mm instead of 6mm).It is a much easier upgrade and probably the right thing to do for such a long Y axis...

Edited 1 time(s). Last edit at 07/02/2017 09:17AM by icefire.

---

Self-sourced Mendelmax 2.0-based Reprap Machine -- Ramps 1.4 & Mega 2560 -- DRV8825 (Z@1A, [email protected], [email protected], E@1A) -- genuine E3D v6 direct setup -- 350W custom silicone heated bed -- ABS 1,75mm -- Marlin 1.1.0-RC7 -- Cura 15.04.6

There are printers using them like the Trinus but it was quite a small machine. Bigger belts might help but I'm lead to believe we are not pushing gt2 belt anywhere it's capabilities with 3d printing. Maybe mount your y axis on linear rails?

Edited 1 time(s). Last edit at 07/02/2017 09:18AM by DjDemonD.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

I used to habe this IGUS linear rail: [www.igus.eu]
However, I decided to replace it and go for much cheaper 10mm shafts and LM10UU linear bearings because the movement along the IGUS rail was not particularly smooth. But I still think there may be oscillation of the whole Y carriage due to the elasticity of the belt itself.. I am thinking of a way to test this theory...

---

Self-sourced Mendelmax 2.0-based Reprap Machine -- Ramps 1.4 & Mega 2560 -- DRV8825 (Z@1A, [email protected], [email protected], E@1A) -- genuine E3D v6 direct setup -- 350W custom silicone heated bed -- ABS 1,75mm -- Marlin 1.1.0-RC7 -- Cura 15.04.6

Quote
icefire
All in all, I guess that printers with moving beds will always have the problem to some extent.

I have another idea: How about using a leadscrew for the Y axis instead of the GT2 belt? My heated bed is 320mm long in Y and the physical length of the axis is 420mm. In addition to that the Y carriage is pretty heavy so maybe the belt is just too elastic for this kind of weight. Maybe a leadscrew will eliminate the issue as it has no elasticity.

Any thoughts on this?

One of my printers has a ball screw in the Y axis, installed to reduce the ringing I was getting when it was belt driven. It definitely improved the ringing but the noise level is so high I had to leave the machine at the makerspace because my wife complained about it. It's driven by a 425 oz-in NEMA-23 motor that has it's own DSP driver and a 32V power supply. Maximum reliable print and motion speed is limited to about 40 mm/sec with that machine because there's a nasty mechanical resonance at about 50 mm/sec that causes the Y axis to skip steps if it hits 50 mm/sec even briefly, such as when printing a circle. Otherwise I think it could print much faster. IRIC the lead of the ball screw is 10 mm/rev.

NEMA-23 motors vibrate quite a bit, even with high microstepping (64:1, IRIC), and the vibration is very efficiently coupled through the ball screw to the bed plate, hence the noise. I've tried antivibration motor mounts, etc. all to no avail. I should probably try a lower torque motor- that might push the resonance up to a higher speed...

---

Ultra MegaMax Dominator 3D printer: [drmrehorst.blogspot.com]

In a Galaxy far away, the War against Ripples continues.

Most are eaten, stop the war, stop these babies coming home in goodie bags.

Quote
the_digital_dentist
One of my printers has a ball screw in the Y axis, installed to reduce the ringing I was getting when it was belt driven. It definitely improved the ringing but the noise level is so high I had to leave the machine at the makerspace because my wife complained about it. It's driven by a 425 oz-in NEMA-23 motor that has it's own DSP driver and a 32V power supply. Maximum reliable print and motion speed is limited to about 40 mm/sec with that machine because there's a nasty mechanical resonance at about 50 mm/sec that causes the Y axis to skip steps if it hits 50 mm/sec even briefly, such as when printing a circle. Otherwise I think it could print much faster. IRIC the lead of the ball screw is 10 mm/rev.

NEMA-23 motors vibrate quite a bit, even with high microstepping (64:1, IRIC), and the vibration is very efficiently coupled through the ball screw to the bed plate, hence the noise. I've tried antivibration motor mounts, etc. all to no avail. I should probably try a lower torque motor- that might push the resonance up to a higher speed...

Ok, screw that (pun not intended )

If we return to the timing belt concept, I´d like to report that did the following test:
- A dial gauge mounted on the printer frame so that the travelling distance in Y can be measured
- stepper motors engaged so that the motor holds the Y carriage and the axis cannot be moved by hand
- I tried rocking the heated bed forwards and backwards by hand (unfortuntely I cannot measure the force I have applied but it was not much) and I found out that there is +/- 0,25 mm elasticity in each direction.
- Next step: I held the motor shaft in place with a pair of pliers to eliminate the possibility that the the holding torque of the motor is not enough and the elasticity comes from there. Result: No change in the elasticity values (I did my best to rock the bed with the same intencity).

This means that the elasticity of the Y carriage does come from the elasticity of the timing belt. So using a wider belt (15mm) should, at least in theory, improve the ripples problem.

Maybe it is worth a try....

---

Self-sourced Mendelmax 2.0-based Reprap Machine -- Ramps 1.4 & Mega 2560 -- DRV8825 (Z@1A, [email protected], [email protected], E@1A) -- genuine E3D v6 direct setup -- 350W custom silicone heated bed -- ABS 1,75mm -- Marlin 1.1.0-RC7 -- Cura 15.04.6

You could use the steel/PU (white) belt its much less elastic but you have to use at least 20t pulleys preferably larger, or the steel wires fatigue and then it becomes rather a lot more elastic.

Edited 2 time(s). Last edit at 07/03/2017 10:34AM by DjDemonD.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions