Welcome to the SeeMeCNC Community

Hi all,

All delta printers I saw until now use the three poles to linearly move one end of the arms using driving belt or fishing line.

I had a look on the net for delta machines used in the industry.

All industrial delta bots I have seen until now have a quite different design, with motors on top, and no travelling trolley of sorts, no belt.
Instead, the motors drive a forearm, elbowed to an arm attached to the effector.

I mean, google for "industrial delta robot", and you'll probably understand what I mean.

I did not say all robots are like this or that, but it seems to be a general truth, from what I have seen so far.

Is there a reason for this ? I understand for the industrial choice of removing belts, and trolleys.
How about the 3D printer designers' choice of keeping them ? Is it a question of cost ? of technology ? of patents ?

If you search for the firepick delta, you will see examples of people struggling to calibrate the thing properly. I believe the radial (as opposed to linear) delta robots also may have more complex math and stricter requirements of precise measurements.

Also, there is no (good) open-source firmware that I know of (unless the firepick is OS firmware?) that has radial delta implementation in it.

Those things are built for speed only. They have big AC servo motors that allow them to go fast while keeping decent positioning accuracy, but they still aren't as accurate as our deltas. Plus, they need a rigid, precise top base for the motors and arms.

I think one of the main problems with the reprap version is that the ARMS ARE TOO DAMN THIN! You use 8mm CF tubes that are like half a meter long, and expect to get rigidity?! Plus the rod ends and connecting pieces are flimsy thin walled plastic. The tubes are rough cut to length, and nothing is as precise as it needs to be.



It shouldn't be hard to make one that works with decent accuracy in a reprap way, but I don't see any advantages to it over our linear tower deltas.

I have done industrial work with this robot.
https://www.youtube.com/watch?v=vtAEIKJLHGw
[youtube]http://www.youtube.com/watch?v=vtAEIKJLHGw[/youtube]

The task was the pick - visually inspect component - place of smd components on to pcb board on a conveyor.

They do move fast, and they are very rigid. At those speeds our deltas would be wobbling all over. The robot did have 1 extra degree of freedom, the end effector could rotate. This was achieved with a shaft going from the top center to the end effector which looked a bit like a 4th arm.

They have servos with encoders on top, they do not use stepper motors. Unfortunately I do not know the resolution of the encoders.

Bot nailed it, the FirePick Delta moves that way. Here's a picture from their website.

[img]http://static1.squarespace.com/static/5 ... ormat=500w[/img]

Thank you guys for your replies.

The firepick is a nice looking machine I did not know about.

I am good with my RMax, but I believe this kind of design is sturdier. I mean, this delta design has less moving parts, so I guess less trouble.

Now, I understand it is harder to calibrate it in the first place, but I do see some advantages to it, in time.

Thanks to all for sharing your experiences, and thoughts.

It looks like you would be limited in the print height with the firepick, also.

Yeah, you do not have unlimited Z.

Unlimited Z ? Hey ! Hot end mounted on quadcopter.... layer fan included !

Still some progress to make on stability and battery life, but this is surely the future of 3D printing.

With added benefits of unlimited x and y

Z would still not be unlimited with a quadrocopter. It will eventually run out of atmosphere through which to fly. Instead, mount it on a rocket! Skycrane to the rescue!

More seriously, it looks like the limit on those firepick machines would be equal to the diameter of the gear(upper arm) , thus in order to expand them, you end up seeing massive weight/mass increases (Build height increases linearly with gear diameter (arm length), while gear (arm) volume skyrockets in proportion), whereas a linear delta just sees linear increases on most components. I imagine that once you have 2 foot in diameter gears, they would be rather less portable, more dangerous, and not so fast or accurate.

I got my wires crossed between the firepick and some other similar deltas that used full gears for a moment, excuse me. But the basic idea remains the same. Those arms must be structural to hold their entire mass, the mass of the lower arms, the hotend and effector, and presumably remain fast and accurate. That's a tall order. They will also protrude sideways at some point, greatly increasing the envelope that would need to be enclosed. But it does look cool.

I think I'd prefer the linear design in most cases, as we already need the structural framing for the entire thing. Might as well use it. It also makes it easier to add a counterweight to the slides, as they only move linearly, thus making payload mass increases easier. And cheapskates are pretty much a solved problem now, thanks to SeemeCNC and similar.

The biggest issues with the rotational delta design is making the arms stiff enough and getting enough resolution out of the steppers, they have to be geared down to increase the resolution which impacts speed.
Industrial robots just don't use steppers.
The linear deltas we use seem like a better fit to me given the technologies that are practical given the price point.

Having said all of that as much as I love my MAX, I'm still not convinced that for printing delta designs offer much of anything in the way of an advantage over a good cartesian design and cartesian printers are much easier to calibrate. The deltas biggest win over cartesian designs is vertical movement speed, and it's just not a big deal for printing.

The original delta robots used the overhead design, as seen in the modern FirePick. That was the way they were all designed since the original patent in the '80s. As pick-and-place machines, their performance is unimpeachable. Where a Cartesian machine moving in the XY plane is limited to using two motors at a time, a delta machine necessarily uses all three, meaning that you get a higher built-in top speed for whatever motors you use.

Also, I know that linear deltas can be made with ridiculously low part count. Deltas are generally "supposed" to be easier/quicker to build than comparable Cartesian machines. If you look at the Trick Laser MAX METAL frame, it's way less complicated than a regular Rostock MAX frame. Part count is FAR lower. With Cartesians, I guess it depends on whether you're using X and Y gantries, or X gantry and Y bed, or CoreXY, or whatever. (I was never a fan of "moving bed" printers.)

There is experimental support for this type of delta printer in Smoothieware. They call it a "rotational delta," if I'm not mistaken. The FirePick is mentioned explicitly in the source comments. The math is not terribly hard. You compute one axis (arm position) at a time, and if all three are calculated correctly, the effector is in the right place. In other words, while calculating the position of axis A, you don't have to worry about axes B or C.

The trick with any delta robot (rotational - or linear, like ours) is always the joints on the carriages and effector. SeeMeCNC has some simple U-joints. Trick Laser and others use Traxxas ball joints. Neither kind of joint is perfect - there will be some play, which is why it's necessary to hold the arms together with rubberbands. Even then, you have to figure the plastic part of the joint is losing its surface a few molecules at a time when the robot is in motion. I've had two sets of Trick Laser arms and they both developed some play after thousands of hours of use. What I would really like to know is what joints would be used on an industrial robot, and whether these could be had for a price reasonable to a home user. I would like to see something that A) has close to zero play, B) stays that way even after thousands of hours on the Hobbs, and C) can bear a significant load without making it wear down faster. I don't know of any joints that fit the bill for that, but it would be nice. I've played with mounting extruders to the arms, and thought about the idea of having them on a platform sitting on "stilts" above the effector platform, etc. but it always comes down to the question of whether it's worth it to wear out my joints prematurely to do it. The best solution I'm aware of is to use large neodymium magnets, but I don't know how even those would perform if each arm had to carry the weight of a stepper motor in addition to the weight on the effector.

I suspect that the linear delta (Rostock) model was implemented at least partially because figuring out a sufficiently rigid double-jointed arm, with enough resolution to operate from cheap 1.8-degree stepper motors, is daunting. If you drive the arms directly, you lose TONS of positional accuracy; your 200 steps/rev motor would have perhaps 50-70 steps to represent its ENTIRE travel range. That's not nearly enough, so you'd have to use gears, or geared steppers. Geared steppers are common now, but in 2009 or whenever the Rostock was designed, I'm guessing the whole RepRap world was stuck on 1.8-degree motors without any gear reduction at all. There is also the leverage to consider, and a sufficiently long double-jointed arm will have a fair amount of leverage against the motor.

You could get a decent build height if the upper and lower arm segments were of a similar length. The FirePick's short upper arms and long lower arms don't give you such a huge Z envelope, but are marvelous for lightning-quick moves, as the effect of rotating the upper arm a little bit is amplified greatly by the length of the lower arm. The higher up the upper arms can swing, the taller the envelope.

I think the rotational type of delta is plausible for printing. We would need a few upgrades and design changes:


1) All rotational axes based on back-to-back preloaded ball bearings. No play, tolerances <<0.001, rigid, long operating life.

2) Thicker CF tube arms with aluminum connectors and joint fasteners.

3) 400 step/rev NEMA 23 steppers coupled to the wheels with a preloaded, backlash free setup.

In the area of smaller CNC milling machines in the machine shop world they mostly have abandoned stepper motor design for Servos and in expensive machines they went all servo a long time ago.

Yes the hobby market CNC machines such as those from http://www.sherline.com/index.htmlstill use steppers but once you get to Bridgeport size machines only the cheap conversion packages use steppers anymore..dealing with dropped steps require good encoders or interface with high quality glass scales and once you have those anyway AC servos make more sense.

why have steppers held on so long in the 3D printing world? Is it entirely a cost thing?

Cost and inertia largely. There's also some desire to minimize weight, bulk, ETC. we have a Tormach in the shop on campus that uses a stepper for the 4th axis fixture. It's a 23, and a long one, so plenty of torque. But compare that to a Haas 4th axis. You don't pick that up. You use a lift. The difference in size is pretty impressive. Then look at the servo's the Haas's use for main axis motion. They are the size of gallon jugs some of them. Servo's for say a Fanuc are smaller, but in many cases slower. You also have to design a system that actually supports closed loop control, which would make it either incompatible with all other controllers and motors, or very expensive as you now have to support two different control methods and hardware setups.

Printer's are still in their infancy, and honestly, the lack of closed loop control doesn't hurt us much. I've got a 60W laser cutter. Not a piece of consumer crap by a long stretch of the imagination. It uses the same steppers as the Rostock Max. Seriously. Exactly the same model number and manufacturer.

Also, I looked at DC servo's. I found one with a Nm rating of .11, for only 120$. That's about the cheapest I could find on a quick look. Uses a 42MM frame size, and would need substantial gearing to drive the sort of load we deal with. The high end of torques on DC servo's I found was a whopping 6.3 Nm
A standard stepper motor generates .43 Nms. If you want a Servo with the same torque, without gearing, then you end up paying 240$, and it's substantially larger. 57 MM frame size. And it also has only velocity encoding. If you buy a servo, you have to buy a gearbox, and it has to be a high-precision, preloaded one. Otherwise the backlash from the gearing kills any accuracy gains. Most also appear to be rated to use higher voltage DC than we use. 24V's is the low end. 48 appears to be more common. I found a servo that almost looks reasonable. High speed, high enough torque, gearbox packaged with it, 24Vs. But it still lacks a position encoder (I understand why, those buggers can be expensive), and is 309$s. Doubles the cost of the printer to install them for each tower. And they are huge. 42 MM frame size, and 5.4 inches long.

There is next to no load on a 3d printer. Mills, router tables, etc, need to deal with cutting forces that vary wildly. 3D printers experience almost no load, and no changing load either (unless we have a moving bed). The servos would be a great idea for extrusion control, and also smoother axis movement. The feedback loop isn't really all that enticing to me.

Yeah for most 3d printers, NEMA 17 and NEMA 23 stepper motors work fine. We are limited in speed by lack of machine rigidity and hotend back-pressure anyway. Plus, you can get NEMA 17 steppers for like 9 dollars each with the stepper drivers costing only a few dollars more on pre-made control boards. Can't get much cheaper.


Even on small CNC machines, like Sherline and Taig, steppers are fine because you need a lot of driving force at a relatively slow speed. The spindle motor is what does the work.

I've been thinking a lot lately about using low KV outrunner RC motors for position control. The ESC's they use are either meant for constant power output or constant RPM, depending on the firmware. I think the firmware could be rewritten or modified for position control, since the motors are brushless and already use back EMF or optos to detect position. Low cost hobby RC motors are the only thing with a decent wattage per dollar ratio.

I have a Tormach PCNC1100 CNC machine with a 4th axis and it uses steppers although these are PolyPhase (also called 3 phase) steppers. This is a Series II machine upgraded to a Series III in many areas. Traverse rates are 120 IPM on X/Y and 90 IPM on Z (Head weights `500#) and lost steps ARE NOT an issue.

There is nothing wrong or inadequate with stepper motors. I have retrofit a lot of very large machine tools (X travel 60 feet - Y travel 30 feet - Z travel 9 feet). Long ago the only large servo motors were hydraulic, then came DC Servos, then AC Servos (PMW DC Motors) and now Digital Servos. It is interesting to note that the Red Cap Fanuc and others like the Digital AB and Yaskawa Servos can in fact be Step and Direction Drives with closed loop positioning between the motor and drive.

The areas where Steppers fall down are, 1) High Torque at High Speed, 2) High Torque Requirements (think foot/pounds rather that inch/pounds) 3) Current Draw At Rest. None of these are really an issue with what we do. As far as a closed loop system, this is a control and feedback issue and can be applied to either steppers or servos.

Stepper motors get a bad rep, and closed loop positioning is greatly misunderstood.

Closed loop positioning on a stepper would in most cases be useless, if you are dropping steps, you are doing so because the torque available to the stepper could not overcome the resisting force (for whatever reason) simply retrying the same move has absolutely no guarantee that it would then succeed in fact in most cases it would repeatedly fail because the fundamental mechanical problem hasn't been fixed. Servo drivers apply increased current to increase the force based on some PWM tuning, Stepper motors just don't work like that.

If you want to knock steppers in the printer space, you can knock the crappy drivers, or the fact that almost all the printer electronics drive them from way to little voltage.

Some stepper drivers have missed step detection, but the pin that does that isn't hooked up to the microcontroller so it's ignored. The upcoming Smoothieboard 2 will have those pins connected. AFAIK, that's all you need for closed-loop. You could also put an encoder on the stepper, if you wanted.

I don't think missed steps are all that common, unless the current is set too low or the printer is being run at excessive feedrate. I noticed that my 0.9-degree steppers will run backwards if I ask them to move too quickly, e.g. set the speed multiplier to 250% and try to home the printer.

I run my entire fly reel manufacturing business with mills and lathes CNC'd with stepper motors. I routinely cut circles on the mills and these have to be precise to .002" with no issues with dropped steps at all. I don't push my machines to the limits of speeds though and I think that has a lot to do with not having an issue. The simplicity and low cost for the performance can't be beat for machines in this size range. My homebrew CNC mills (2 of them) are about the size of a Tormach PCNC1100 and my lathe is a 10x22 machine. I agree, I think steppers get a bad rap.

MACH 3, FTW!!!!

I have found this discussion most illuminating and thank all you folks who have set my misconceptions straight on the subject.

in my own workshop i never made the move to convert my Bridgeport partly due to the cost involved for things like ball screws to replace the lead screws and the fact that almost every thing i do is one off stuff with the occasional 4 or 5 off and once in a blue moon a 12 off of something so my shop is frozen in the 1960s with the exception of frequency controlled 3 phase spindle controls and adding the Heidenhain Positip 855 to my Bridgeport I bought it about 15or 16 years ago used with the 4 glass scales it uses for a very low price at an auction and the way it is set up to take a programed or learned set of moves and present them on the screen for me to repeat has made this very useful for occasional jobs where i have to make a few copies of an item.

but as a result i learned the Positip method of machine code and not G Code which would be far more useful now that i am learning new skills here

for the curious this old system is details in the programing section of this PDF manual for the 855
http://www.atechauthority.com/pdf/pt855_manual.pdf