Posts made by Catalin_RO

I was among the first users of Duet on a CNC machine over four years ago. Back than I had a genuine, first version, Workbee, with several years of using GRBL on another low cost CNC. While Duet was lacking features of GRBL that I was already used with, the old Pololu DRV8825 drivers were not up to the task and I decided not to spend to much on proper Leadshine drivers. So Duet looked perfect as a hardware platform.

With CamBam for the CAM part, configuring it to generate a GCode that was understood by Duet back then was OK (G2 and G3 were not even planned back then, not to mention the work coordinates!). I wrote on the forum about the issues that I have found, I complained about some of the missing features, and things slowly got better.

These days the Duet boards are a good choice, depending on the CNC you want to control.

3 years later problems started to creep in. Being an Electronics engineer, I started digging. It all came down to the on-board drivers. Randomly, only at powering on, they reported disconnected and/or shorted steppers. I checked all connections, I replaced connectors, I ended up replacing cables - no luck! And, worse, sometimes I was not able to use the CNC for a whole day. Once started, all went flawlessly! I was not the only one with the symptom, just search the forum for "phase disconnected". There was no clear solution to the problem!

My ten cents - even if properly cooled, using the on-board drivers at 2.4A (my steppers are rated for 3A) for hours at a time is affecting them over time. The higher power drivers on the Duet 3 board should be better. But there is another problem - the supply voltage! You can't get proper performance from the steppers on a larger CNC without using at least 36V as supply. Even the new Duet 3 boards are limited to 32V.

So I decided to put the Duet board to storage and moved to PlanetCNC MK3 with Leadshine drivers. The bill was slightly higher, but not hugely higher (less than double!). And it comes with a lot of CNC specific bells and whistles.

@joaquin_suave You can get this one - https://ro.mouser.com/ProductDetail/Cosel/PJMA1000F-36?qs=DRkmTr78QARflml570Qxjw%3D%3D - while announced for 36V, it can be adjusted from 30.8V to 40.8V. Unfortunately 32V is not that usual and 20A is anything but low current...

@zapta Quite interesting. Those might be suitable for axis 4 and 5 in a prosumer CNC...

@zapta said in Why not brushless motors in direct drive extruders?:

This is probably a matter of demand. With sufficient demand, this can be a much lower cost IC. The question is, is it useful for large scale applications?

Demand builds up when there is a real need for such a solution. I see a need for it in completely different applications, but not in a 3D printer, not even in a prosumer one. In a professional one, maybe!

Has anyone complained about the steppers limiting the performance or the quality of the 3D printers on this forum? I have not personally checked, but from already 4 years of browsing I don't recall anything significant. Of course, I'm not discussing about faulty steppers or super cheap steppers assembled in a barn in China (by the pigs grown up for feeding the family over the next year) or poorly chosen (wrong size for the job) ones. There are very good quality steppers at very decent prices, most of them manufactured in China. And they are not even difficult to find.

So with a relatively low demand, combined with the high current involved in those drivers (120A peak current! those transistors are anything but cheap!) I don't see the solution getting significantly cheaper. Overall you must also factor in the high current power supply. 24V at 60A is almost 1.5kW peak power for one of these motors. While the average required power is significantly lower, the PSU should be ready to handle those peak currents. Look at this just for reference - https://www.onlinecomponents.com/en/mean-well-usa/rsp200024-43879729.html.

As for recognized professional solutions, check this servo with integrated driver - https://www.sorotec.de/shop/JMC-Servo-Motor-with-integrated-driver-100-Watt---36-Volt---3000-1-min.html. Significant torque with very high speed as 36V and 6A peak current.

@zapta said in Why not brushless motors in direct drive extruders?:

I am not an expert but isn't this what gears do, trading between torque and speed?

Of course, but those gears and their supporting plates start adding to the weight. And the whole discussion started from the 30g BLDC motor.

@rjenkinsgb said in Why not brushless motors in direct drive extruders?:

They both use permanent magnet rotors and wound stators.
(Steppers do not have plain iron cores - if you turn one, you can feel the rotor magnet jumping from alignment with one stator pole to the next).

Unless they are variable reluctance ones!

@zapta Extruding is all about torque at relative low RPM. This is what I see for those really interesting solutions:

• Maximum torque 1.99Nm or 3.86Nm, but that comes with a peak current of no less than 65A!!!
• A big note - "*Note that torque and current ratings are with Extremely good forced air cooling"
• RPM at no load - 5760RPM or 8640RPM.

The torque I get, at much lower speeds of course, from a rather standard NEMA 23 at 2.4A and no need for extra cooling.

And the pricing... the driver is a merely 179EUR. You get 3 or even 4 decent Leadshine drivers for that amount!

Again, that is exactly what I have stated... it is a solution, bot not for extruding the filament! I would see it more likely for the gantry movement where, in a 3D printer, the required torque is very low but the speed is high. But, still, it might not justify the extra costs...

@pertti A stepper motor is the most simple model of brushless motor. The ones that you think of, as used in RC models, have 3 phases (they have 3 coils) and are designed for a very high power/weight ratio and quite high RPMs. Also they produce a lot of heat that must be somehow removed - not a real problem in any fast moving model/drone.

The steppers are simpler as they use just two phases (they have just two coils), are designed for much lower RPM and much higher torque and are cheaper as they need just iron (high quality magnets are crazy expensive since the whole EV craziness - but this off topic!). Not to mention the significant precision. Also, the overall required power is significantly lower and also is the produced heat.

While you don't see why filament extrusion requires precision, just think about a very finely detailed object that needs printing. That implies a lot of quite dramatic direction changes, so the extruder precision is important in order not to over- or under-extrude. If you follow the various topics on this forum, you will see a serious debate on various extruding rate estimation/models. If you consider the 1.75mm filament and a 0.3mm nozzle, for every 1mm of printer head movement you need to extrude just under 0.03mm of filament.

What you are suggesting is to take a motor that literally has only much fewer steps - twice the number of magnets on the rotor, design a microstepping controller for it and use it instead of the well proven steppers. The ESC in any RC model is, in the end, just an oscillator with 3 outputs with 120° between their phases and a frequency matched to the required RPM. It doesn't have all the blows and whistles in any cheap stepper driver!

So, while not impossible, it sounds more like reinventing the wheel. And based on my own experience I bet that the final result, for the same performance, will be more complex, heavier and a lot more expensive!

P.S. Please don't get me wrong! I'm not against new solutions. Even at work I'm known for looking first of all for cons in any technical suggestion and I have my own share of "exploring the uncharted territories" activities. It's just that I'm used to look at all aspects of any alternative technical suggestion. Some of them I gladly embrace myself and help testing/developing!

@deckingman From the picture, the printer we are discussing about has moving steppers. So the heat dissipation would happen only if the metallic plates would be quite massive. A simple solution would be to have the Y axes steppers fixed in the rear or the profiles and the belt looped over a pulley or a large idle wheel in the front. That simplifies cabling, hugely improves steppers cooling and reduces the gantry weight.

As for static electricity breaking the steppers, never! While static electricity may cause you small shocks, due to the relatively high voltage, the current is almost impossible to measure due to its very low value and too short time. The stepper coils are designed to handle relatively high currents when compared to the current in a static discharge.

But the high voltage in a static discharge is a guarantee for "success" when we consider the stepper driver. While it may not break the driver, it may briefly (or permanently, but then you blame the driver for breaking the stepper!) open all the semiconductor junctions in it, practically getting a large current into the stepper, with no current limiting logic. That may break the stepper! But I never saw that happening!

Static electricity buildup requires a multitude of conditions to be met simultaneously. I don't see how those could be met in a 3D printer supplied, usually, with 24V, unless we are discussing of a large amount of hot air moving inside the printer so that it would get really dry. Normal air humidity reduces the chance of static electricity buildup.

And even with Aluminum plates, there is no proper electric connection between the elements, at least not while using the V-slot wheels. Even with other solutions of linear bearings there is no guarantee as the V-slot profiles are usually anodized, so they have a non-conductive protective layer.

@michaelr123 If you rotate the profile by 90° you will start having issues on the Z axis. If you want something really right, just replace the profile with a beefier one. With minimal changes, the 4040 should be significantly less susceptible to bending. Or, even simpler, attach another 4020 or 20*20 profile to your existing one. As you have nothing running on the lateral slots, just drill 5.5mm holes in through the center of the extra profile, larger through the V-slot so that the screw head will pass through, and secure the extra profile with slot nuts to the existing profile.

For a 1m profile use 5 or 6 screws evenly distributed over its length. Minimal effort for a significant rigidity gain.

You could stick to 3D printed brackets for now, but go for something like PETG and make them a lot thicker. It also helps to design them with guides fort the rails to slide in, so you get the required 90° in the corners a lot easier.

For the stepper metal plates, you should be able to find shops that provide laser cut Aluminium or steel plates. 6-8mm aluminum or 4-6mm stainless should do the job a lot better than the plastic ones. Thicker plates are needed only for milling CNCs. Those companies usually need just a DXF or a technical drawing of the plate.

@michaelr123 In order to properly tighten the belts you need a solid frame. The plastic plates will not provide the required mechanical strength when you properly tighten the belts. Go for thick Aluminum (8-10mm) or even steel.

Also, when you tension the belts, use a caliper to insure that the travel is the same on both Y axes. If your 40*20 profiles are perfectly equal, just measure the distance from the edge at both ends after a controlled move (same number of steps, same travel distance). It is a lot simpler than measuring the actual tension in the belts.

From your picture I would guess a 1m 40*20 profile for the Y axis. That might be too long for proper belts tension, the profile might slightly bend sideways - not by much, but enough to create artifacts in the middle area. The eShapeOko profiles had embedded V-rail and that gave a better rigidity.

Also, from my own experience, after using both V-rail and V-slot, the V-rail is a lot more precise. The massive Aluminum corners are more resilient to dents and small bends while the V-slot margins are quite fragile. When still playing with the original WorkBee, I had artifacts in the microns range, visible when doing surface milling. Those artifacts looked like a checkered table. After replacing the V-wheels on the Y axes with supported linear rails, the artifacts remained only on the X axis.

I think, overall, you all have a rather wrong image! While I'm using only a CNC for now, I had my belt-based CNC years ago - eShapeOko, 750mm*750mm (https://amberspyglass.co.uk/store/eshapeoko-cnc-milling-machine-mechanical-kit.html). It could extremely easy do 10000mm/min rapids (that is over 160mm/s) with a 2kg router (including its mount). Of course, dual Y so the gantry is kept squared after properly aligning it upon power up. Back then I was using GRBL shield with DRB8825, so not as powerful as the Duet boards.

With CoreXY I presume the most critical aspect is keeping the gantry squared! In the end you have 3 degrees of freedom controlled with just two actuators. They behave as well as a proper dual Y-axis solution only as long as the parallel axis are close enough to overcome any tendency of the gantry to flex/rotate. A large CoreXY gantry, in order to be perfectly stable, would be much heavier than a proper dual Y-axis solution.

You might not like the situation, but here is a simple explanation:

In order to maximize the build volume, all the linear rails in a 3D printer have just one bearing on each of them. The linear bearings, no matter which type, have a very inconvenient rule, in this situation - you must have at least two bearings on the rail, significantly spaced apart, in order to compensate for any rotation tendency. One bearing, while it seems to be impossible to rotate on the linear rail, it does! And when it happens, it gets stuck to the point of breaking things.

So there are two possibilities:

• Beef-up the gantry, especially by using two linear bearings on each rail and compromise the build area, but also increasing the profile size
• Stick to dual Y-axis and have a moving stepper for the X-axis and a light gantry

Overall, I think that sticking two dual Y-axis is simpler and easier to scale up.

Your problem exists as well in the router CNC world as well. If you look at all those Chinese 6040 models, they have the same issue, even if their linear bearings used on Y-axis are 3-4 times longer than the ones that you normally use on 3D printers. They still have 3 degrees of freedom with two actuators. That is why most decent routers have dual Y-axis!

@jrow Sorry to hear that!

Normally the filter, if connected the other way around, should be OK. The schematic has no component that would be in a critical condition if connected the wrong way, just that its characteristic would be slightly different. Among other things, I'm involved with a company that manufactures TEMPEST equipment (if you don't know what it means, just consider it EMI for paranoids!) and line filters are the first line of defense.

The VFD should have included a proper short protection toward the spindle, unless the spindle is "just slightly" broken. If you have a multimeter I would check the resistance between the 3 power wires and then the resistance between each of them and the ground. The 3 resistances between the power wires should be equal and pretty low. The resistances between the power wires and the ground should be really high, in the MOhm range. If one of the power wires has a significantly lower resistance to ground, there is your short circuit.

When using 3-phase spindles you don't really need a braided power cable. The beauty of AC engines is that they don't produce a lot of electrical noise unless they have a problem.

Now on the Meanwell clone power supply... It's like going for a Leadshine clone for your stepper! It is similar and covers most possible problems, but at some point it might hit you really hard! I have seen and I still see a lot of Chinese clones in various low cost, and not only, equipment. Usually they have a lot of empty places on the PCB, usually on the filtering side, either AC or DC.

Just for reference, I have used a Vigortronix 48V/10A PSU for a project (Vigortronix is anything but cheap!!!) where it was located in an enclosure on top of a building. 10 months later it was busted. In a similar location I have a Meanwell doing the same job for over 5 years.

@jrow Indeed, you have installed the filter in the wrong direction. The 3-pins end is for the load, as indicated on its label (though pretty uncommon!). Also, don't expect much from it as the common mode inductor and the X capacitor are pretty small. Better go for something like this one - https://www.conrad.com/p/schaffner-fn2090-16-06-emi-filter-250-v-ac-16-a-4-mh-l-x-w-x-h-1135-x-575-x-454-mm-1-pcs-554095. While it is significantly expensive, from my own experience it makes a huge difference.

But this filter just prevents electrical noise from getting back into the supply lines. You also need to insure that the VFD is properly grounded and also check that the spindle is really grounded. I had to make myself the ground connection into my spindle as the Chinese ones usually are not properly grounded.

On my CNC I have a Huanyang 1.5kW VFD inverter and a 24V power supply from Meanwell for the Duet. Never had any EMI problems and I have no line filter on the inverter. For connecting the spindle to the VFD I use a 4-core Lapp cable. The 4 cores are slightly twisted for maximum flexibility and that also helps a little bit on the electrical noise side.

@o_lampe As long as the GCode file can be opened with various viewers (I have myself tested it with a very old, mostly GRBL tool - Candle and with the PlanetCNC application in demo mode) and it looks OK it can be only a problem with RRF.

But, also, I wouldn't rule out a mechanical issue even if disabling G2/G3 commands in the postprocessor gives correct results, especially because it is a belt-based machine. While splitting in the postprocessor the G2/G3 commands in small straight lines might seem identical to what RRF is doing, the simple fact that those lines have different lengths may change the behaviour of the machine.

I have myself started years ago with a ShapeOko clone and then moved to the original WorkBee. Both pretty decent machines if all possible problems where considered when generating the GCode - both machines where technically backlash free (impossible to determine/measure backlash in the belts/screws), but with other crazy "degrees of freedom", all related to the wheels solution.

As a simple comparison, with both machines I broke quite a few 3mm cutters when going just 500mm/min at 0.2mm depth. With the QueenBee I did machining at 1200mm/min, 0.5mm depth with a 2mm cutter!

Hi @psychotik2k3,

If you dig further into that old thread, you will see that the jerk could be increased over time. Right now I still have it set to 400mm/min for X and Y, much lower for Z and 4th axis. But this parameter is used only when changing the movement direction on each axis. When machining/printing circles the jerk speed should normally not be used, or it shouldn't really matter, as the circles come up on each axis as a very natural movement, especially if the circle is large.

Have you considered looking for possible backlash sources? From my own experience that is the most overlooked cause for such problems. And it must be checked in all components.

I had my own ShapeOko clone above 5 years ago, but bought as a complete kit from an UK company. The delivered parts where of very good quality and I could get very precise results on Aluminum even if it was using belts and my milling motor was a beefy Proxxon weighing almost 3kg.

I also have access to a low cost Chinese engraving laser, with a flimsy frame and small steppers. That one exhibits problems similar to yours and we have tracked it down to the pulleys on the steppers. Somehow they are not perfectly mating with the belts and introduce some backlash, no matter the feed rate.

Even the 0.1..0.3 tolerance on your milled Aluminum parts indicate more of a mechanical issue. Even with my old ShapeOko clone I could get under 0.05mm precision. With my WorkBee I had to spend quite some time to properly determine the best steps/mm values as it is much easier to manufacture a close to perfect belt than to machine a leadscrew. Now I have to do the same for the upgrade - QueenBee (btw, pretty rigid with the 800W spindle).

So, if I were you, I would start milling small squares, like 50mm*50mm, and check their size, edges and corners. And that should be done at least in 9 places so you can check all extremities and central areas of the axes. if you have around also a decent machinist square, you can also insure indirectly that the machine is perfectly squared. The backlash should also be visible when machining squares, usually by not having perfect 90° corners or straight edges.

The Z-axis lack of rigidity may have two components. One is a rotation around the X axis due to the flimsiness (can't call it otherwise!) of the V-slot in that setup, and another one is rotation around the X axis plate (grab the spindle perpendicularly to the X axis, from the front, and try rotating it CW and CCW - you might find some wobble). The rotation around the X axis make impossible to machine thick items with perfectly vertical edges and tend to easily brake small end-mills (had my share of broken end-mills with the WorkBee), but it can be ignored for thin items (3-4mm thick!). The rotation around the X axis plate, though, produces a lot interesting artifacts, including some backlash-like. Also, this rotation around the X axis plate produces thick pieces that are smaller at the top than at the bottom on the X axis.

So, do some tests on acetal, better if 10-15mm thick, and check for all angles and sizes. From there a lot of things can be understood and it will be easier to find the problem. Check the resulting pieces after machining in each position of the OX. When you find a problem, just stop and ask as it would be a pity to waste the material and the time.

@dc42 It always happens when trying to home the machine after a software reset. Right now I can easily reproduce it:

• power on the whole machine
• do a homing cycle - everything is OK during this run if none of the end-stops is triggered
• do an emergency reset through the web interface
• try to home again the machine - the errors start popping up

I can't blame it any of the connectors because the errors do not come up after the initial power on. Also, if instead of trying to do a complete homing sequence I try to home a specific axis, the errors will come up for that axis. Even more, as I have a dual-Y setup, when I try homing the Y axis, the errors come up for both drivers/steppers.

Once the machine is homed, I can used for days with no errors!

More to that: the machine is a QueenBee mechanics (similar to WorkBee), with high torque steppers left from the original WorkBee - 3A, 4mH, 1.2Ohm.

More to this...

Just upgraded to 3.4.0beta7, mostly out of curiosity. Immediately after the update everything looked OK. I homed the machine several times without a glitch.

So I decided to do a minor change in the config.g file. I saved it, soft rebooted the machine and there it was! The problem hit me again...

In the end I was able to get rid of the issue after a quick series of power cycles. As previously observed, once properly started, everything worked flawlessly.

As this time I had no homing switch triggered that brings me to a very poor conclusion - somehow the Duet board degraded over time and in certain situations it starts to wrongly indicate motor related errors. I do not suspect issues related to its relocation after upgrading the machine as I have moved it along with its support structure.

Overheating is not an issue for sure. Last time I have extensively used the machine about 6 months ago, with environment temperatures not exceeding +30°C. These days, the room where it is located is at a rather cool 21-22°C and there is a fan continuously cooling the whole board. No matter what I do, the CPU never reports more than 31°C. I don't suspect any of the drivers getting too hot, especially while doing these simple homing tests.

Is there any way of getting around the phase disconnected check in the Trinamic drivers? Right now it seems to be mostly a bogus error more than anything else. Or should I consider getting a new controller?

@theolodian The reset loop is something that I have also seen and I properly understood immediately. The problem is that if the board is powered on with any homing/limit switch triggered, it doesn't enter the reset loop and, instead, it falsely reports issues with the steppers.

While I intend to make a lot more clever homing scripts, separating the homing and limits switches electrically (different inputs) and not enabling the triggers for an axis until homes, the problem is still there...

@phaedrux The homing switches are also used as safety limits, thus there is a small retraction before resetting the coordinates. In series with the homing switches, there is another set of switches that are really used as safety limits.

The trigger based solution is the most simple that insures an abrupt stop if any switch is activated when it shouldn't.