Duet Wifi

[BTP]

I've been doing some research on trying to resolve vertical banding and it seems the Duet controller is the go to upgrade to eliminate this. What is it about this controller that makes the difference when compared to the Rambo board?

[Xenocrates]

The first obvious feature in all the duet boards is that they have non-segmented moves, meaning motion is continuously calculated, so you can achieve something much closer to an actual circle (As the motors have distinct steps, it isn't perfect, but it is below human resolution for the most part). The second big one is that the Duet wifi has much better stepper drivers, which rather than having jumpers set the mode pretty much permanently, can be intelligently controlled, including during a print, to change the microstepping modes and otherwise have improved resolution (IE, 16X microstepping on the Rambo, but up to 256X microstepping on the Duet wifi, a potential gain of 16 times the resolution (Although torque is minimal at that step rate), meaning that rather than 80 steps/mm, it can be 1280, or sub micron resolution on each tower directly.)

[mhackney]

Have realistic expectations though, many of the STLs on thingiverse are absolute crap - low polygon counts. You will never get a great print out of an inferior STL file. If you are designing your own models, learn how to tesselate properly-I have a post on my blog about how to get the best results.

[BTP]

Thanks guys.

[adarcher]

The first obvious feature in all the duet boards is that they have non-segmented moves, meaning motion is continuously calculated, so you can achieve something much closer to an actual circle (As the motors have distinct steps, it isn't perfect, but it is below human resolution for the most part). The second big one is that the Duet wifi has much better stepper drivers, which rather than having jumpers set the mode pretty much permanently, can be intelligently controlled, including during a print, to change the microstepping modes and otherwise have improved resolution (IE, 16X microstepping on the Rambo, but up to 256X microstepping on the Duet wifi, a potential gain of 16 times the resolution (Although torque is minimal at that step rate), meaning that rather than 80 steps/mm, it can be 1280, or sub micron resolution on each tower directly.)

What's the method of getting the better circles? Does the firmware detect curves and recalculate them? I would think the slicer controls the movement of the head by giving coordinates to move to--a straight line.

Also, I thought that changing the stepping mid print was a nono. Have they improved the math/timing on this? Theoretically it should be possible, but I was under the impression that the firmware author was against it.

The benefits I saw in the upgrade were: improved math capabilities over the 8bit rambo, and the higher resolution stepping. As well as all the firmware/software that came with it.

[Xenocrates]

The better circles come out of having the calculations done for each step, to ensure that you're getting all the possible precision, rather than each move being cut up into small lines, and assumed to be close enough. I don't have a whole lot of the math to describe it, but while the slicer controls precision somewhat (I can define an arc, or a series of very tiny moves in the G-code of a part, and some will not complete them very well due to the segmentation), if you were on a Cartesian, then segmentation would never be an issue. On a delta, the "straight" line may be distorted by the need to continuously update the carriage positions to get as close as possible. Same with circles. At a certain point, the segmentation is equivalent to the stepper resolution, at least for now, so it becomes academic, however the Rambo, unlike the smoothieboard and some others that also have segmented steps, does not have the processor power to calculate small enough segments at a high enough speed for it to go un-noticed. You can look at Mhackney's sample parts for some of this, which already shows the difference in the Duet .8.5, where the drivers are of the same quality. The major difference is in the quality of the path calculations in the firmware.

As far as I know, the stepping settings in the firmware should not be changed mid print. However, the drivers have the capacity to do interpolation and dynamic step rates on their own, to reduce noise or smooth motion. However, given DC42, I have no doubt that sooner or later, if there is any real benefit to be had from it, that in his pursuit of perfection, the firmware will support it better, and the capacity to do so remains, well supported or not.