Upgrading to ARM and 0.9 Steppers - Vertical Banding Issue

[mhackney]

I do A LOT of CNC milling with 200 step per rev steppers with 16 micro steps. Both with direct drive and 1-1 timing belts driving the axis. I've never seen any artifacts on my milled parts that I would attribute to stepper resolution. And a milling machine has to be MUCH more rigid than a 3D printer. The electronics and software for these machines are more "sophisticated" so the the calculations may be carried to more precision but at the end of the day, .001" resolution is the norm on these machines (talking small production and hobbyist CMNC mills). My fly fishing reel parts are milled and then finished with a very light cut with an ultrasharp end mill. The surface finish from this is like a mirror in aluminum and brass. No artifacts like this at all.

The lack of rigidity and lightness of these machines may very well be contributing to this.

[Generic Default]

I do A LOT of CNC milling with 200 step per rev steppers with 16 micro steps. Both with direct drive and 1-1 timing belts driving the axis. I've never seen any artifacts on my milled parts that I would attribute to stepper resolution. And a milling machine has to be MUCH more rigid than a 3D printer. The electronics and software for these machines are more "sophisticated" so the the calculations may be carried to more precision but at the end of the day, .001" resolution is the norm on these machines (talking small production and hobbyist CMNC mills). My fly fishing reel parts are milled and then finished with a very light cut with an ultrasharp end mill. The surface finish from this is like a mirror in aluminum and brass. No artifacts like this at all.

The lack of rigidity and lightness of these machines may very well be contributing to this.

Yep, you nailed it with that. CNC machines have much heavier carriages that don't respond to the frequencies that are created by the half-step jumps I described in my previous post. Our delta printers have lighter carriages than just about any cartesian printers, and yet most people are using 3/8 inch tubes or nylon truss arms! It guess it's all about frequency response for us. Also keep in mind that even if your machine was extremely jittery your surface would still have a minimum radius the same as your end mill, since it's a spinning cutter. Printer nozzles can make tiny, relatively sharp corners on inside pocket geometries.

You're talking about the surface finish on the *side of a milled part, right? That will mirror the cutting edge of the end mill (which is precision ground). The bottom of a milled pocket doesn't reflect like a mirror, which is why you [probably] fly cut the flat surfaces for a nice looking reflective surface.


And about the high end Stratysis Dimension machine I was talking about earlier. It has the carriage mass and similar construction to a small CNC mill, but the acceleration that the engineers decided to give it is way too high for the high-mass moderate-rigidity construction of the machine. I would expect better design for something made by a team of professional engineers; hobbyists like us have already passed them in mechanical design in my opinion. At least that's what I think when comparing my rostock printed parts to my Stratasys printed parts.

[mhackney]

Yes, surface finish on the side of a milled part - which is cut by the cutting edge of the end mill. I used that as an example since it would reflect artifacts in the X-Y movement plane with a fixed Z similar to our printers.

Maybe we could devise a few hypotheses and tests for them and "divide and conquer" or replicate each other's work to try to determine the root cause? Sort of a group debugging exercise.

[626Pilot]

I'm just waiting to see what happens with 0.9 degree steppers. I'd buy four of them if I knew someone else already proved it was worthwhile.

[Generic Default]

Once I make my own printer from scratch I'll be sure to buy the 400 step/rev steppers. I don't see any advantages to the 200 step/rev steppers in this application. If we were using leadscrews like on a cartesian CNC machine, the 200's might have an advantage with torque at higher speed. For now we can't reduce the vertical waves much without more positioning resolution.


Also if you print a feature that's a vertically slanted face (such as a 45 degree overhang) you'll notice that the vertical waves become diagonal waves; this is because the half-step jump is being phase shifted by the angle of the overhang. It can make some cool looking effects like interference patterns if you design your part for it.

And mhackney, one thing your CNC mill doesn't have to deal with is that it doesn't interpolate all three axes at once to maintain Z height. Any backlash in a delta printer will cause minor Z lifting in a repeating pattern from discrete steps and stick-slip friction. This can also cause bad surface finishes on printed parts, but it looks different (and more severe) compared to the Z banding that xnaron brought up with this thread. I guess the goal all of us share is that we want an ultra smooth surface finish on our printed parts, up the level you know with your CNC mill.
I like your idea of debugging for this problem. Most of us with Rostocks have noticed the imperfections in print quality, and in order to advance the technology we need to develop open source solutions for these technical problems.

[xnaron]

I've got the 0.9 steppers installed along with the ramps-fd running the ARM Arduino DUE. I just finished a test print with the printer. I used the same repetier config. The vertical banding is definitely still there but a lot fainter. It is closer to the banding I have on my cartesians. I need to run some more tests.

[xnaron]

I'm playing with different values for the printing segments. I tried 400 and it isn't as nice as 180 (default) was. I tried 800 and ended up with some stuttering. I tried setting it for 70 and I think it may actually look better. This is on a cylinder printed as a vase. I still need to do a lot more tests before I can really say how much of an improvement this may be. At this point I'm not hating it.

[mhackney]

Wow, that's a lot of advancement! I'd love to see the pics of the setup.

So, it seems like more processing power helps but doesn't eliminate. Perhaps the culprit is the algorithm?

cheers,
Michael

[xnaron]

Wow, that's a lot of advancement! I'd love to see the pics of the setup.

So, it seems like more processing power helps but doesn't eliminate. Perhaps the culprit is the algorithm?

cheers,
Michael

I'll make a youtube video tomorrow of the setup. So for the results I'd say cylinders are on par with a cartesian. As for straight lines on axis the cartesian wins. They straight lines are markedly improved now though but the vertical banding is still visible. I tried segments of 70,180 and 400 on a 2 loop cube sliced in vase mode with kiss. The 180 segment one looks the best followed by 70 and then 400. It is a big improvement to the 1.8 steppers. I think that the majority of the improvement is due to the increased resolution of the 0.9 deg steppers. I don't think the ARM processor is really adding anything here as the firmware is the same. I think what we really need is to see some delta firmware written for these arm controllers or at least optimizations of the current flavors.

[xnaron]

Here are some pics

cartesian on left; delta on right

cartesian on left; delta on right

[xnaron]

An added benefit of the 0.9 steppers is that they are quieter than the 1.8 steppers. I didn't install the noise dampers with these. I think there is a slight quality penalty with them on so I opted to leave them off. I was concerned about going back to the noise as they really made it quiet. It was a happy surprise to find these 0.9 steppers to be quieter than the 1.8 stock steppers. They aren't silent like the dampers but they are a happy compromise.

[626Pilot]

Did you try upgrading the extruder to a 0.9 degree motor?

[xnaron]

[youtube]http://www.youtube.com/watch?v=3sMw_UAM994[/youtube]

[xnaron]

Did you try upgrading the extruder to a 0.9 degree motor?

No. I am running a geared stepper. I do want to try a 0.9 on a wades variant. I need to pick up a hobbed gear though before I can do that.

[enggmaug]

for .9 degrees steppers, should you buy unipolar stepper or bipollar ? what voltage ?

[xnaron]

for .9 degrees steppers, should you buy unipolar stepper or bipollar ? what voltage ?

I believe these are the ones I have http://www.kysanelectronics.com/Product ... ordID=6875" onclick="window.open(this.href);return false;

However I wouldn't recommend upgrading to 0.9 steppers without moving to the ARM and some compatible shield such as RAMPS-FD or RADDS. The atmel arduino mega will not have enough horsepower to drive the 0.9 steppers at whatever top speed you use now. As the steppers require 2x the steps your max speed will be divided by 2.

Right now I am tweaking the Repetier ARM board in concert with others including the repetier developers. Lucky for us the dev also has a DELTA. Once I come back with a good working config I will post the details.

I wouldn't advise anyone buy the Rev 1A version of the RAMPS-FD on ebay made by getech. This is because it is an earlier version that has bugs. Some of them are safety bugs. There is a problem with the way the fet circuit is designed for heaters. Right now if the ARM board fails it is possible for the heaters to turn on and of course this would be unregulated by the thermistors and could start a fire.

[xnaron]

[youtube]http://www.youtube.com/watch?v=L_stokiQRZ8[/youtube]

[youtube]http://www.youtube.com/watch?v=wdiY4GKug-I[/youtube]

[626Pilot]

Psst... it's not doo, it's doo-ay. It's an eye-talian word!

[xnaron]

Psst... it's not doo, it's doo-ay. It's an eye-talian word!

lol...thanks and will call that out on the next vids.

[mhackney]

Xnaron, I don't know if you know this, I just found out today, but there is a special #define EXACT_DELTA_MOVES that improves the delta calculations. It was originally created for ARM boards but .91.7 Repetier has experimental support for AVR. I've enabled it (just add #define EXACT_DELTA_MOVES to your Configuration.h file) and testing now. Unfortunately, I'm printing a part in black PLA now which hides everything! I'll move to a translucent PLA for the next print that might show more.

[xnaron]

Xnaron, I don't know if you know this, I just found out today, but there is a special #define EXACT_DELTA_MOVES that improves the delta calculations. It was originally created for ARM boards but .91.7 Repetier has experimental support for AVR. I've enabled it (just add #define EXACT_DELTA_MOVES to your Configuration.h file) and testing now. Unfortunately, I'm printing a part in black PLA now which hides everything! I'll move to a translucent PLA for the next print that might show more.

I'm pretty sure this applies to the AVR version of Repetier and not the ARM version I am using. I sent Roland a PM asking him just to be sure.

I am currently running 4x the resolution I was with the stock rmax. I doubled the resolution when I wen from 1.9 to 0.9/deg per step steppers. I doubled it again when I we from 1/16 to 1/32 stepping on my allegro drv8825 stepper drivers. I went from 80 steps per mm to 320. The latest bleeding edge version of repetier is amazing for ARM. I have DELTA_SEGMENTS_PER_SECOND_PRINT set to 600 and DELTA_SEGMENTS_PER_SECOND_MOVE set to 200. I am able to print at the same speeds I was before the upgrade. The performance is simply amazing. I have reduced the vertical banding so it is barely noticeable. I think any artifacts now are due to mechanical issues. Yesterday I bought Simplify3d slicing software. I am getting some of the nicest prints I have ever made.

[mhackney]

According to Roland on the repeater github issues list the EXACT_DELTA_MOVES subsegment computation was originally written for the ARM version. Have you heard back from him?

[xnaron]

According to Roland on the repeater github issues list the EXACT_DELTA_MOVES subsegment computation was originally written for the ARM version. Have you heard back from him?

Yes I heard back from him. He said Due always has it enabled. He made a switch for avr because he wasn't sure the cpu could handle it.

Brendin

[mhackney]

Great. I've got something like 8 hrs with it enabled now and 6 prints with no performance issues. But, I have been printing in black so haven't been able to validate effect on part quality/appearance. Hope to do some of that today.

cheers,
Michael

[xnaron]

[img]http://i57.tinypic.com/2hwen9l.jpg[/img]