Is it worth upgrarding to a 32bit controller like a Smoothie or Replicape?
Posted by MrMeeeseeeks
|
Re: Is it worth upgrarding to a 32bit controller like a Smoothie or Replicape? September 17, 2016 08:13AM |
For the last 18 months I've been printing and modding/building machines I've been quite against the 32 bit approach due mainly to cost. Also I'd not had any real problems with ramps/arduino.
My newest machine has duet wifi and its a massive improvement. It's a bit like driving around in a Ford fiesta saying "this car is enough and gets me from a to b" then getting a tesla model S and realising there is a big difference.
That'll learn me.
Model s next
(seriously though if cost is a big factor and you're building a printer on a shoestring get the least dodgy ramps board you can from a supplier in your country, who can swap it if it breaks. If you're onto your 2nd, 3rd or 4th machine and quality, reliability and ease of use matter more than cost then then it's a no brainer).
Edited 1 time(s). Last edit at 09/17/2016 08:14AM by DjDemonD.
Simon Khoury
Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions
My newest machine has duet wifi and its a massive improvement. It's a bit like driving around in a Ford fiesta saying "this car is enough and gets me from a to b" then getting a tesla model S and realising there is a big difference.
That'll learn me.
Model s next
(seriously though if cost is a big factor and you're building a printer on a shoestring get the least dodgy ramps board you can from a supplier in your country, who can swap it if it breaks. If you're onto your 2nd, 3rd or 4th machine and quality, reliability and ease of use matter more than cost then then it's a no brainer).
Edited 1 time(s). Last edit at 09/17/2016 08:14AM by DjDemonD.
Simon Khoury
Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions
|
Re: Is it worth upgrarding to a 32bit controller like a Smoothie or Replicape? September 28, 2016 03:42PM |
At least with the old Marlin the segment ends are precise to ±0.5*microstep. I.e. rounding to the nearest microstep is the only error at segment ends. In the center of a segment, the error is at most additional 6 µm for 3 mm segments. This additional 6 µm can cause that it will be rounded to a different microstep than if there would be no segmentation error at all. Therefore I said at most 1 microstep error in the center.
How this is possible? I know you do not do segmentation therefore each carriage position is precise to ±0.5*microstep. You cannot easily achieve more precision because our stepper motor drivers are discrete. With the discrete stepper drivers, the steppers either move by a microstep or do not move at all. There is nothing in between. This is definitely true for small speeds. With higher speeds, you could use the fact that static stepper motor moment is finite (and rather low for bigger microstepping) and that the rotor/carriage/platform/hotend have some inertia and motor coils have some inductance and belts are rather elastic. Therefore the actuall speed of head movement will vary with something like an offset sinusoid and not constantly (or according to the curve from delta-cartesian transformation). The more quickly you print the more "flat" the sinusoid will be. And the more precise intermediate positions. Provided that your stepper inductance is not too high for given speed and power source voltage. This is a bit better than the ideal discrete movement (which would happen if there would not be any inertia in the system). Still, the intermediate head positions between cube corners will not be perfect. Of course they are not perfect even in your pictures, because if they would be perfect then we could not see the moire in the pictures.
For precise intermediate positions, you need analog current control to the stepper motor phases. Without analog current control to the motors you will get rounding to the nearest microstep, at each position, regardless whether it is at the cube corners or in the middle of the wall. This is more true the slower you print. If it would not be true you would not see any difference on the cube walls between these two pictures. You would see a difference only at the cube corners.
Quote
dc42
When my printer prints a cube, simultaneous rounding of the position to the nearest microstep for all three motors only happens at the corners. Between the corners, the microsteps for each motor are generated independently at precisely (to within a few microseconds) the right times.
How this is possible? I know you do not do segmentation therefore each carriage position is precise to ±0.5*microstep. You cannot easily achieve more precision because our stepper motor drivers are discrete. With the discrete stepper drivers, the steppers either move by a microstep or do not move at all. There is nothing in between. This is definitely true for small speeds. With higher speeds, you could use the fact that static stepper motor moment is finite (and rather low for bigger microstepping) and that the rotor/carriage/platform/hotend have some inertia and motor coils have some inductance and belts are rather elastic. Therefore the actuall speed of head movement will vary with something like an offset sinusoid and not constantly (or according to the curve from delta-cartesian transformation). The more quickly you print the more "flat" the sinusoid will be. And the more precise intermediate positions. Provided that your stepper inductance is not too high for given speed and power source voltage. This is a bit better than the ideal discrete movement (which would happen if there would not be any inertia in the system). Still, the intermediate head positions between cube corners will not be perfect. Of course they are not perfect even in your pictures, because if they would be perfect then we could not see the moire in the pictures.
For precise intermediate positions, you need analog current control to the stepper motor phases. Without analog current control to the motors you will get rounding to the nearest microstep, at each position, regardless whether it is at the cube corners or in the middle of the wall. This is more true the slower you print. If it would not be true you would not see any difference on the cube walls between these two pictures. You would see a difference only at the cube corners.
|
Re: Is it worth upgrarding to a 32bit controller like a Smoothie or Replicape? September 28, 2016 06:06PM |
I cannot easily do the test with Repetier since I do not have model for delta segmentation error of Repetier. That is the reason I was interested in the pictures. To see if it looks like what MrBas suspects. In other words I wanted to see something without effort 
And when I changed on Marlin from SPS 200 to SPS 90 I did not notice a quality change. But I was not really looking for it either. Somebody else reported that he did not notice a quality change when lowering SPS from 200 down too. Maybe if we would look hard enough we could notice it. My computations showed that the difference between head position with SPS 200 and SPS 90 should be at most one microstep for my printer geometry and my print speed. And I decided that one microstep difference does not matter since other errors are much bigger at the speeds/accelerations I print. I think my biggest problem is the belt elasticity. When looking at the objects, I guess it contributes about 0.1 mm error. It is clearly visible around sharp edges of the printed objects. One does not see this error on round objects since accelerations are low when movement changes smoothly. I guess this is the reason people like to print vases with deltas. It is the most easy object to print nicely. It is smooth (therefore elasticity problems are not visible) and it does not require a lot of filament retracts (therefore plastic deposition errors due to bowden will not appear).
Another problem (which dwarfs segmentation errors) is different shrinking of plastics as it cools depending on the horizontal area size. Heat chamber should help here. But I do not have it.
My point is that compared to these errors, one microstep additional error because of delta segmentation is not even on my radar. I could be on my radar if I would print slowly and with low accelerations. But that is not the direction I want to go.
And when I changed on Marlin from SPS 200 to SPS 90 I did not notice a quality change. But I was not really looking for it either. Somebody else reported that he did not notice a quality change when lowering SPS from 200 down too. Maybe if we would look hard enough we could notice it. My computations showed that the difference between head position with SPS 200 and SPS 90 should be at most one microstep for my printer geometry and my print speed. And I decided that one microstep difference does not matter since other errors are much bigger at the speeds/accelerations I print. I think my biggest problem is the belt elasticity. When looking at the objects, I guess it contributes about 0.1 mm error. It is clearly visible around sharp edges of the printed objects. One does not see this error on round objects since accelerations are low when movement changes smoothly. I guess this is the reason people like to print vases with deltas. It is the most easy object to print nicely. It is smooth (therefore elasticity problems are not visible) and it does not require a lot of filament retracts (therefore plastic deposition errors due to bowden will not appear).
Another problem (which dwarfs segmentation errors) is different shrinking of plastics as it cools depending on the horizontal area size. Heat chamber should help here. But I do not have it.
My point is that compared to these errors, one microstep additional error because of delta segmentation is not even on my radar. I could be on my radar if I would print slowly and with low accelerations. But that is not the direction I want to go.
Sorry, only registered users may post in this forum.