3D Printer Engineering

[John ShepherdParticipant @johnshepherd38883]

Following on from an exchange of views I had with Neil in another thread about how many of the popular 3D printers are engineered, is there any interest in a new thread about the subject?

In a thread about cheap printers, I complained about what I consider to be poor practice such as bearings held in place with cable ties, poor rigidity, 'all threaded' rod used for construction and for lead screws etc. and outlined some of the modifications I have made.

Neil expressed the view that the loads and demands on a 3D printer are a lot less than on machine tools and there is no need to employ the same engineering standards.

My stance is that there are many variables in 3D printing and it is good practice to reduce as many as possible. Besides, I am more at ease using a better engineered printer just as I would be when using any other well made machine.

I realise that a lot of printers are supplied as kits and many users do not have engineering experience to build a complex design or the machining facilities to improve them, but if that route gets them into 3D printing that is good.

There are a lot of modifications to 3D printers featured on the internet and many of them are ill founded, so perhaps it would be good to get some input from the wealth of knowledge on this forum?

If there is an appetite for it, I will be happy to put my head above the parapet and put my thoughts on the firing line.

John

[John ShepherdParticipant @johnshepherd38883]

[JasonBModerator @jasonb]

Maybe you could answer the question I raised in the other thread to kick things off.

Neil Commented

"Going from a pitch of 0.75mm to 4mm (typical for ballscrews sold for 3D printers) is reducing the vertical precision by a factor of more than five"

Jason asked:

"Would some of that lost precision be regained by the better accuracy of the ball screw/nut combination compared with a bit of studding and whatever is used as the nut?

I can see that with a finer pitch you get better resolution but if that is say +-0.01 for studding vs say +-0.001 of a ballscrew you have gained as much as you have lost"

[SillyOldDufferModerator @sillyoldduffer]

Go for it John, but be aware that there are two schools of thought on 'good practice'. One is that engineering should seek perfection, or at least the best that can be achieved. The other is that 'good practice' means making items that are fit for purpose at minimum cost. The second is what most professional engineers do for a living, and they know that chasing "quality" for it's own sake is an egregious sin.

Of the two 'value engineering' is far harder to get right than the perfectionist kind. Was it George Stephenson who said "An engineer is a man who does for a pound what any fool can do for a guinea."? Many problems are only difficult when working to a budget.

In hobby-land most 3D printers are aggressively engineered down to a price. If they weren't the average Joe couldn't afford one. So of course they can be improved, but is it worth it? You don't want to waste time and money fixing things that are already 'good enough'. On the other hand, the designer may have gone too far or made mistakes: if the design and build are skimped then it might well be worth improving it.

A risk of upgrading inexpensive equipment is that you introduce new problems and end up polishing a turd. If I were to boost the engine of my very ordinary car with NOX, I'd soon find that the clutch and brakes needed upgrading as well. After seizing the engine on my way to the shops I might twig an extra radiator is needed to keep the oil cool. Several thousand pounds later, I might think my money would have been better spent on buying a performance car in the first place. You have to know when to stop.

That said, quite a few people enjoy owning good tools, and they enjoy improving what they own. Even though the aimless pursuit of quality isn't 'good practice' in professional engineering terms, I think it's completely respectable in a hobby. So don't be put off – I'm interested in what you have to say.

Dave

[BazyleParticipant @bazyle]

If you think the leadscrew has an error shouldn't it be possible to compensate. After all isn't that how some of the bed levelling processes work?

People used to buy a basic car and add features and trim even a radio when they weren't standard because they couldn't afford the whole cost at once but could progressively. So the same should be true for printers.
Particularly in our hobby I would expect to replace plastic bits with metal. Maybe picking a common design like the Prusa i3 and providing drawings of key bits to improve would be a valid topic for MEW.

[Ian PParticipant @ianp]

Posted by JasonB on 16/11/2018 11:01:01:

Maybe you could answer the question I raised in the other thread to kick things off.

Jason asked:

"Would some of that lost precision be regained by the better accuracy of the ball screw/nut combination compared with a bit of studding and whatever is used as the nut?

A straight answer to that question must be yes.

Surely though is not as simple as that. Without defining why the precision was lost it would be difficult to say whether a coarser ballscrew would give a better result.

I spent 2 hours this morning deciding what to buy and this thread came up whilst my finger was hovering a 3d printer purchase button! Before I pressed it I thought I would just have another read of the several 3D threads on this forum hoping to learn some fact that I had missed before that would help tip my choice in the right direction.

I do not expect machine tool quality build in a the sub £300 printer but I would like the basic engineering of the printer to look at least as if it had been well thought out. Understandably the main effort for the manufacturers is probably deployed in cost reduction. I prefer linear shaft over wheel in groove type guides and an aluminium rather than wood or plastic frame construction. One model I looked at has the vertical axis leadscrews mounted on the stepper spindle with a machined ali (helical slit) flexible coupling. Since this type of coupling is designed to have axial compliance it seems a poor design choice if, as I understand it, nozzle height is very important. I know the weight that the leadscrew is carrying is going to be pretty constant so the axial length of the coupling will not change much but to me the coupling has added an element of uncertainty into the nozzle height.

There are some parts I want to make by 3D printing and I know as I have a working setup will realise there are many others that I will want to print. From the property descriptions of the various filaments PLA will be fine initially for the parts I want so I suppose I could get the simplest/cheapest printer to gain experience on but I would then like whatever cheap printer I bought to be capable of being improved by myself.

As of today I am veering towards the 3D Factory units that Neil recommended, only delivery time (well, my patience) might rule as I would like to get one within the next few days.

Ian P

[Andy CarruthersParticipant @andycarruthers33275]

More power to your elbow John

I am extensively modifying a 3D printer right now, one of the minor aims is to reduce noise so replacing bearings with IGUS equivalent and more serious aim is to significantly increase print volume in all axis

One feature to prevent constant setup is auto-bed levelling, a real time saver, perhaps one of your posts will address this

[Neil WyattModerator @neilwyatt]

Perhaps it's worth listing some of the modifications that DO seem to be worthwhile (I haven't done those marked with a *):

Getting rid of low quality connectors and replace with direct solder or better connectors.

Using separate boards* to control high-powered heaters (reliability/safety).

Insulating under the hot bed (faster response, reduced energy consumption – but enclosed printers used for cold-air vulnerable materials may benefit from the increased 'heat leak from an uninsulated bed).

Things that either reduce the flexibility of the frame* OR its resonant frequency (the irony that placing a kilo of filament on top of a frame can actually improve performance by dropping the resonant frequency).

Using nylon/delrin linear bearings* instead of linear ball bearings (longer service life/greater reliability – allegedly)

Improved filament guides in extruders (allow simpler filament change/use of flexible filament)

Larger steppers (within reason reduces risk of lost steps etc. my machine came with larger than normal)

Automatic levelling (just makes life easier, the net time saving of auto-levelling every print versus one manual levelling a every week or two is moot)

Improved support for filament reels (convenience)

Improved support and cable runs (reliability, safety)

Borosilicate glass for hot bed (much more durable)

It's also worth bearing in mind that (for fused filament machines) the limiting factors on print quality are ultimately nozzle diameter and filament quality. Vertical steps less than 0.02mm are impractical as it's hard to reliably extrude such low amounts of filament. The usual nozzle size is 0.4mm, although 0.2mm is practical (if much slower, about four times slower), this clearly imposes a limit on the value of extreme horizontal resolutions; my experience is that when tuning a printer worrying about less than 0.05mm with a 0.4mm nozzle is not worth worrying about. These figure translate to surface roughness.

Experience shows that three things have the most marked effects on tuning print quality:

• Making sure you detect and replace a worn nozzle as soon as print quality starts to degrade.
• Getting the right compensation values for first and subsequent layers to avoid 'elephant's foot' and get accurate X-Y dimensions.
• Getting the best temperature for your filament.

[John ShepherdParticipant @johnshepherd38883]

Ok Here goes.

I will feed in just one or two mods/issues at a time to hopefully keep the thread focused.

My printer is based on a Prusa i3 and uses a Duet WiFi board (perhaps more on that later). It runs on 24v, has a laser bed level compensation system, heated bed and a glass print surface. ( I will add an image later)

The issue that raised some response was using an Acme leadscrew (or a ball screw) instead of threaded rod as used on the Z axis in the original Prusa design. It was suggested that using a Ball screw (and I suppose an Acme thread), threw away resolution. My printer came with Acme threads on so can’t comment on if there is any improvement in print quality but it seems to make sense to use properly machined components made for the job rather than studding that because of how it is made may not be straight or have a great deal of dimensional accuracy.

At the risk of getting the math wrong, I won’t go into the calculations, but my z axis parameter is 400 full steps for 1mm movement. The TMC 2660 driver on the Duet has 256 micro stepping and I can get Z movements of 0.05mm with repeatability by using the function on the interface. I don’t think that is any sort of compromise. Smaller value ‘Baby step’ offsets can also be made on the fly and can be useful in making adjustments to squash down a first layer on difficult prints, it is also useful for setting up the nozzle height above the board, but I have not found any other uses for it.

Connected with the Z axis is the laser module that is used at the start of each print to adjust one of the two Z axis motors to eliminate any tilt of the bed and then to measure the relative height of each corner of the bed so that software corrections can be applied if needed. It also acts as an ‘endstop’ and allows precise setting of the nozzle to bed height.

That all for now, but there is more!

[Neil WyattModerator @neilwyatt]

It's an interesting debate, John these are my musings.

The M8 allthread and 1.8 degree steppers on my machine have a basic step resolution of 0.004mm or 250 steps per mm. I can't quite see how you get 400 full steps per mm using a ball screw with a conventional stepper?

I think the basic problem with ballscrews is that they are more vulnerable to lost motion as the stepper typically needs about five times more torque, although allowing for the increased efficiency of ballscrews the difference may be a bit smaller.

Having looked very closely into what steppers actually deliver under load for driving telescope mounts, I remain sceptical about the real world accuracy of all but the most modest levels of microstepping.

Having said all that, I suspect that there is no practical difference in the results obtained – although I came across someone who found long-pitch (8mm) ballscrews both hard to drive accurately and also prone to drive backwards and drop the z-carriage whenever the power was removed from the steppers.

As for accuracy.. actually the stainless steel allthread used from 3D printers and better quality ballscrews are both produced by thread rolling and should both have similar high levels of inherent accuracy, certainly more than enough for 3D printing. It's essentially only the thread form that differs. I will agree that allthread may wear out faster than a ballscrew, but it should last years and years…

Nut accuracy is irrelevant, as they should be floating and all the loads should be unidirectional as you observed.

I should stress that I nearly hit the 'buy' button on ballscrews, it was only after going back and assessing the actual impact on my prints that i decided it wasn't worth my while.

There may be issues with the spiral cut connectors, but I can find no problems with them if assembled properly.

A confession, I fitted upper bearings to my z-axis rods. This was just to make me feel better, as even if the rod wobbled 10mm side to side at the top the maximum vertical error would be 0.04mm In practice even 2mm would be extreme because the nut constrains the actual movement and this would have a vertical component of 0.001mm.

So clearly, I am not immune to the temptation to solve non-existent problems, but does feel more accurate if the tops of the z-axis screws don't waggle around!

Neil

[John HaineParticipant @johnhaine32865]

FWIW…

When I first converted my Super 7 to CNC I wanted to avoid any irrevocable changes to Myford's components. I decided not to use the original X-feed screw because it looked tricky to arrange to drive it and made a new screw using stainless M12 all-thread. Yes, I know that 1.75 mm isn't an obvious thread for a feedscrew but Mach 3 takes care of that. I did have a number of problems with accuracy, but none that I could definitively pin on the thread. I don't personally think one should worry too much about thread pitch accuracy based on this. The mean pitch is likely to be pretty accurate and anyway you can calibrate it.

I did in the end fit a ballscrew on the X-slide because taper turning is a PITA if you have significant backlash. One day I may fit one on the leadscrew…

[John ShepherdParticipant @johnshepherd38883]

As promised, here is my printer:

[John ShepherdParticipant @johnshepherd38883]

Picking up points about the Z axis that MAY make a difference for better or worse.

First re Steps/mm – my M92 Reprap Gcode values are X80 X80 Z400 E417 ; set steps per mm. using an Acme 4 start lead screw

I have seen a lot of comment about fitting bearings on the free end of the lead screws and as I have not experienced much in the way of wobble have followed the example used on my lathe cross slide and other x y tables where the free end is left to float. Perhaps someone could explain the pros and cons of this?

I used a plum connector to connect the leadscrew to the stepper motor, not particularly because it provided better alignment, in fact the axial movement is quite stiff, so I doubt it moves much with the loads applied to it, but because it has a clamping arrangement on the two shafts. My previous coupler was just a straight one held in place with grub screws. I didn’t like the idea of grub screws on shafts with no flats or indentations because of the possible damage that might make disassembly and reassembly difficult. I also tried the flexible couplings with spiral slots in them, but these introduced more wobble and artefacts on the vertical part of prints, probably due to poor manufacture or perhaps design. Again it would be good to have a definitive view on the use of these.

In the past I have tried the sprung anti backlash nuts on the lead screws and these were useless, in any event the weight of the carriage on the Prusa takes care of that. I now use Delrin nuts that are smooth and don’t have much play.

John

[Neil WyattModerator @neilwyatt]

Posted by John Shepherd on 17/11/2018 09:25:07:

First re Steps/mm – my M92 Reprap Gcode values are X80 X80 Z400 E417 ; set steps per mm. using an Acme 4 start lead screw

You printer looks an awful lot prettier than mine!

M92 includes microstepping, so 400 is an 8mm pitch 4-start leadscrew (not ballscrew) with a standard 1.8 degree stepper (200 steps) and 16 step microstepping (3,200 steps/rev). It's exactly 1/10 the resolution of the 0.8mm pitch M5 thread on my machine which has the same number of steps/rev.

A bit academic as 400/mm is more resolution than you need, but I would worry about the 10x reduction in torque affecting fast moves.

Does your z-carriage remain in the up position with power off? I know on bigger printers the steppers can be driven backwards under the weight.

Your couplings look good, Huco couplings might be another alternative.

Neil.

[MuzzerParticipant @muzzer]

Posted by Neil Wyatt on 17/11/2018 13:47:33:

A bit academic as 400/mm is more resolution than you need, but I would worry about the 10x reduction in torque affecting fast moves.

That's a common misconception but you should save your worry for something else. The torque of a synchronous motor (including stepper motors) is proportional to the movement from the commanded position ("positional error&quot. If you generate 10x microsteps, you haven't magically redesigned the motor. Each microstep now represents 1/10 of the fundamental step – and oddly enough a movement of the motor by a 1/10 of a step from the commanded position generates 1/10 of the torque. It has the same torque vs angle behaviour, so it will behave in exactly the same way as the un-microstepped motor, assuming the driver is capable of processing the higher frequency pulse train.

Murray

[Joseph Noci 1Participant @josephnoci1]

The issue of torque in microstepping is often confusing. As Murray indicates, there is no loss of fundamental torque. There is however a loss of incremental torque, during which the motor might not move.

If the load torque plus motor friction and detent torque exceeds the incremental torque of a microstep, successive microsteps will have to be realized until the accumulated torque exceeds the load torque plus the motor friction and detent torque, allowing the rotor to actually move.

Taking a microstep does not mean the motor will actually move. And reversing direction can require a large number of microsteps to get the motor to move.

Incremental torque/microstep as the number of microsteps/full step increases

Microsteps/full step        % holding torque/microstep

1                                    100.00%

2                                    70.71%

4                                    38.27%

8                                    19.51%

16                                  9.80%

32                                  4.91%

64                                  2.45%

128                                1.23%

256                                0.61%

The table shows the significant impact of the incremental torque/microstep as a function of the number of microsteps/full step.

 

Microstepping as a means of improving resolution or accuracy is quite overrated – especially in mechanics that have higher friction of movement, or stiction in the system. What microstepping does do, is make the movement much smoother and assist in overcoming motor ( rotor ) resonances.

 

Joe

 

Edited By Joseph Noci 1 on 17/11/2018 16:24:31

[Martin ConnellyParticipant @martinconnelly55370]

Put hybrid servo steppers on. If you have a 1000 line encoder and microstepping set to 1000 pulses per rev the driver will up the torque until it achieves the required movement. There will also be an error signal available if something is wrong such as hitting hard stops and there is a resulting miss-match between required position and actual position. You would have to consider the cost benefit of a more expensive set up if it is for a hobby not a business.

Martin C

[Neil WyattModerator @neilwyatt]

Posted by Muzzer on 17/11/2018 15:13:52:

Posted by Neil Wyatt on 17/11/2018 13:47:33:

A bit academic as 400/mm is more resolution than you need, but I would worry about the 10x reduction in torque affecting fast moves.

That's a common misconception but you should save your worry for something else. The torque of a synchronous motor (including stepper motors) is proportional to the movement from the commanded position ("positional error". If you generate 10x microsteps, you haven't magically redesigned the motor. Each microstep now represents 1/10 of the fundamental step – and oddly enough a movement of the motor by a 1/10 of a step from the commanded position generates 1/10 of the torque. It has the same torque vs angle behaviour, so it will behave in exactly the same way as the un-microstepped motor, assuming the driver is capable of processing the higher frequency pulse train.

Murray

Nothing to do with microstep ratios.

The leadscrew has 10 times the lead, so to lift a load through a given distance, ten times the torque will be needed.

[Neil WyattModerator @neilwyatt]

In more detail, the 4-start leadscrew has a pitch of 8mm . The allthread is M5 so 0.8mm pitch. From http://www.engineeringtoolbox.com/screw-jack-d_1308.html

The effort force for a screw jack when neglecting friction can be expressed as

F = (Q p) / (2 π R) (1)

where

F = effort force at the end of the arm or handle (lb, N)

Q = weight or load (lb, N)

p = pitch – distance or lead of thread in one turn (in, m)

R = lever-arm radius (in, m)

The torque acting on the screw can be calculated as

T = F R

So the torque required is proportional to Pitch.

Also as Joe demonstrates too much microstepping has a detrimental effect on accuracy (there are ways of addressing this with most drivers, but they aren't implemented on typical 3D printer Polulu drivers as I found when making my telescope driver and had to modify the boards to get smooth movement microstepping at about 1 pulse per second)..

Hybrid steppers would be fine, but are are clearly OTT in this applicationa s long as the printer is set up correctly.

[Neil WyattModerator @neilwyatt]

Posted by Joseph Noci 1 on 17/11/2018 16:21:25:

 

Microstepping as a means of improving resolution or accuracy is quite overrated – especially in mechanics that have higher friction of movement, or stiction in the system. What microstepping does do, is make the movement much smoother and assist in overcoming motor ( rotor ) resonances.

Astrophotography allows us to measure the precision and accuracy of our telescope drives when we are autoguiding. Typically we use 1 or 2 second exposures on a guide star whose location can be determined to a typical accuracy of ~0.1 arc-second or better (about 1/20 of the pixel size in arc seconds).

My HEQ5 mount has about 0.144 arc-seconds per microstep resolution (64 microsteps). Tracking aims to compact polar alignment errors, cone errors in the mount, flexure of the scopes, diffraction effects and pitch errors and other imperfections in the drive train.

If microstepping was 100% accurate, the RMS error ought to be close to the microstep resolution.

In the real world the best long-term tracking accuracy I have achieved is 0.6 arc-seconds RMS and some people claim to push 0.5 arc-seconds on a good night.

Clearly it's fairly close, but there's a noticeable real-world difference between the theoretical and actual performance (it's spending most of its time pointed within an 8-microstep band centred on the right point*), and this is probably largely down to the limits of microstepping (direct drive mounts with encoders do rather better).

Incidentally I usually image at pixel scales of between 0.9 and 2.7 arc-seconds, if the RMS error in each axis is close to the pixel scale you get good results.

In terms of accuracy, that's the equivalent of keeping a moving target the size of a swimming pool in your cross-hairs on the opposite side of the Earth

Neil

[John ShepherdParticipant @johnshepherd38883]

OK moving on from astrophotography and stepper motors, one of the other issues that receives a lot of comment on printers is the standard of wiring and connectors used.

My solution was to move away from the Arduino plus RAMPS controller board that I had used previously and use a Duet3D WiFi board. Not only does this use suitable screw terminal blocks but has Molex KK style connectors that lock in place in one direction only. The board is larger than the RAMPS so that makes room for neater, more reliable, wiring.

I have a lot of admiration for how the RAMPS interfaces with the Arduino but a custom board with several additional features designed for the job, although not without challenges, is a big improvement. Changing over to the Duet was like moving from a ZX81 to a Pentium PC.

The board can be powered by 12 or 24 volts and I decided to opt for 24v. This is a safer option because of the reduced current and the wiring to the heated components can be reduced in size. Consequently, the wiring is easier to terminate and more flexible when it needs to be. Heated beds are normally dual voltage and the cost of replacing the extruder heater and the fans is low. Also, I needed to replace the PSU anyway.

Wiring the print head can be a bit of a challenge with a stepper motor, two fans, a thermistor, and in my case, the IR Z probe. A small PCB and removable multiway connectors made wiring much neater and maintenance easier. The heater uses a separate ‘Deans’ connector often used as a battery connector on model cars etc. though.

Good quality crimping tools made for the connector in use are a must and I have learned the hard way by skimping on these. Some people recommend soldering crimps, but I believe there are good reasons not to do this and prefer to use them as intended, besides if crimp connections are good enough on aircraft, they are good enough for my printer.

Has anyone got any alternative solutions?

John

[J BENNETT 1Participant @jbennett1]

Out of interest I just went to the Factory 3D website. Hovering over the "Buy now" icon shows an Ebay URL. Going there takes you to factory-3d selller listings, but all that is there is a few odds and ends, no sign of machine kits. Is this the right company I am looking at? The only other factory 3d I could find was a company in Bulgaria!

[Neil WyattModerator @neilwyatt]

Posted by J BENNETT 1 on 21/11/2018 14:39:21:

Out of interest I just went to the Factory 3D website. Hovering over the "Buy now" icon shows an Ebay URL. Going there takes you to factory-3d selller listings, but all that is there is a few odds and ends, no sign of machine kits. Is this the right company I am looking at? The only other factory 3d I could find was a company in Bulgaria!

They seem to run out of kits as fast as they can print them. At least (unlike a company I tried to buy off recently) they don't list things as in stock and cancel your order without explanation.

Might be worth emailing them and asking when they are likely to have stock back in?

Neil

[J BENNETT 1Participant @jbennett1]

Thanks for the very prompt reply. I will send them an email, although with my current on-going projects I am not sure if I really need another diversion!!

However, SWMBO, who is keen on arts and craft, mainly painting, drawing, stained and fused glass, saw one on a television programme last night and was very interested. The problem will be that she will want to produce her own designs and I don't relish explaining the learning curve that a CAD package will involve.

[Neil WyattModerator @neilwyatt]

Posted by J BENNETT 1 on 21/11/2018 15:03:20:

Thanks for the very prompt reply. I will send them an email, although with my current on-going projects I am not sure if I really need another diversion!!

However, SWMBO, who is keen on arts and craft, mainly painting, drawing, stained and fused glass, saw one on a television programme last night and was very interested. The problem will be that she will want to produce her own designs and I don't relish explaining the learning curve that a CAD package will involve.

She obviously needs the MEW tutorial series

Neil

[Author]

Posts