This engineer claims 100 grams direct extruder. I want one

Quote
Koko76

Quote
jonnycowboy

Quote
Koko76
Good luck Paul, I believe that you will find "closing the loop" to be more difficult than you think. Modern machine tools still keep servo encoders on the motor shaft instead of the moving table for very good reason, the same one which makes it difficult to close the loop here. Personally I think this is a dead end and I am working to make a lightweight direct solution. I think I can get pretty close to the proposed magic number.

Actually why do we care what speed the extruder motor is moving? What we really are interested in is the quantity of filament extruded (or retracted).

Thus:
Why not use a small motor such as the one linked above: similar to stratasys geared maxon motor and instead of an encoder on the motor, just use the mouse sensor to determine filament movement?

Lots of torque, easy to control, and very lightweight.

Because moving the encoder further away in the chain from the source of motion makes it far harder to control. As I said in the post you quoted, modern machine tools still put the servo encoder on the motor shaft, not post gearbox, and not on the moving table. A simplified version of what will happen is that the motor will hunt back and forth through the lash present in the system to get to the commanded position. It becomes harder to damp, and the max velocity and acceleration go way down.
Servo systems are wonderful things, they can be very accurate and powerful when sized and above all tuned correctly. But getting them to that point is certainly more complicated than dropping in a stepper motor, especially with the power levels we are talking about. A DC servo system will be marginally cheaper but nowhere near as simple or adaptable to different builds, loading and other factors. It's not a worthless endeavor, but I fail to see exactly what problems this solves.

Thanks for your reply Koko.
What this solution "solves" is the magical 100g flying hot end which will never be acheivable with a stepper based solution. But yes a stepper will be easier to integrate.
Servos are not that difficult to control - a good PID and high resolution/accuracy and you're 90% of the way there. The rest is software tuning.

Quote

What this solution "solves" is the magical 100g flying hot end which will never be acheivable with a stepper based solution.

Unless the stepper is at the other end of a bowden. I've never used a bowden arrangement, but it seems lots of people do - what is it about bowden that we are trying to avoid/improve on?

Would it help to use a (chamber) pressure sensor instead of encoder+ filament diameter to control extrusion rate?

With a decoder and a bowden it's still impossible to use flexible filament.

I have this discussion with a friend, telling me servo motors are the one and only option for a CNC, while i am capable of machining within .02mm with my 'ordinary' stepper driven cnc and that number goes down to .01 if i'm using more effort. He's right; closed loop is better in theory, but it's not holy. In the end it's the creativity of the machinist that makes the product from 'good' to 'perfect'.

The closed loop bowden filament feeder is rejected, in my case.

For me, the problem isn't that the wrong amount of filament is being fed (if this is so relevant, you should also have a temperature optimized room, since plastics are rather temperature sensitive, size-wise)

This is feasible with our 'shit mechanism' printers; Pretty awesome if you ask me.

[scottwells3d.tumblr.com]

How about something like this, plus a small interface chip to turn a step train into a PWM signal?


[www.hobbyking.com]

Back of the envelope says that it should have loads of torque, around 10 kg at the teeth of the extruder gear. It's continuous rotation, so should be able to drive the extruder gear directly.

Quote
JamesK

Quote

What this solution "solves" is the magical 100g flying hot end which will never be acheivable with a stepper based solution.

Unless the stepper is at the other end of a bowden. I've never used a bowden arrangement, but it seems lots of people do - what is it about bowden that we are trying to avoid/improve on?

Sorry I was unclear - this is a direct drive hot end / extruder and not for a bowden setup.

Quote
jonnycowboy

Thanks for your reply Koko.
What this solution "solves" is the magical 100g flying hot end which will never be acheivable with a stepper based solution. But yes a stepper will be easier to integrate.
Servos are not that difficult to control - a good PID and high resolution/accuracy and you're 90% of the way there. The rest is software tuning.

You don't just go out and get "a good PID". P, I, and D are terms in an equation used to control motion, you can read more here : [en.wikipedia.org] .Those are what is tuned, in order to optimize the feedback loop such that you have positional accuracy without excessive overshoot. They are tuned to the dynamics of the system in question. Different inertial loads, different motor characteristics, and other variables make it challenging to implement in a way that is friendly to methods of construction which frequently use the cheapest possible thing one can come up with, or simply what is available wherever they happen to be.
You can have as accurate or high resolution encoder as you want in the system, and it won't make a well performing servo system if it's not in the right place so that it's input can be used in the feedback loop. If the encoder is not mounted so that the driven motor can be read directly the motor will hunt back and forth through whatever play or lash is inherent in the system. This destroys motors, especially brushed DC motors. It's also not especially great on gears.
I know that it is possible to have a servo driven system, but I don't see any advantage to doing so. I'd also consider something "solved" if there were details of a specific implementation, or perhaps a working design. I have a 70g pancake stepper sitting here on my desk, I have determined through actual experimentation torque requirements, and it exceeds these. I may not be able to get the filament drive, heater and heatsink under 30g, but I'll be close enough that I don't particularly care.

About splitting the encoder from the motor; You can't compensate structural misses (play, backlash) with PID tuning. You need an intelligence that's capable of determining the difference between the input and the output. and store these differences to make it 'as close as possible' (stil, not ideal).. But this would require both an encoded motor and an encoded reader to determine the difference between both. (still, not ideal)

Quote
to3dornottobe
About splitting the encoder from the motor; You can't compensate structural misses (play, backlash) with PID tuning. You need an intelligence that's capable of determining the difference between the input and the output. and store these differences to make it 'as close as possible' (stil, not ideal).. But this would require both an encoded motor and an encoded reader to determine the difference between both. (still, not ideal)

I understand PIDs well, I use them in my day job. For an extruder however play and backlack won't be critical however since the extruder is almost always running quasi-steady state. In the start/stop motion we don't care if there is some backlash since as soon as we start up again it will take up the backlash.

I agree for a servo motor on the movement axes it would not work so well.

Quote
jonnycowboy

Quote
to3dornottobe
About splitting the encoder from the motor; You can't compensate structural misses (play, backlash) with PID tuning. You need an intelligence that's capable of determining the difference between the input and the output. and store these differences to make it 'as close as possible' (stil, not ideal).. But this would require both an encoded motor and an encoded reader to determine the difference between both. (still, not ideal)

I understand PIDs well, I use them in my day job. For an extruder however play and backlack won't be critical however since the extruder is almost always running quasi-steady state. In the start/stop motion we don't care if there is some backlash since as soon as we start up again it will take up the backlash.

I agree for a servo motor on the movement axes it would not work so well.

I still don't think you understand. Although my opinion of what's critical for and extruder gear differs from yours, what I am referring to is play within the servo loop itself. If there is play between the motors operation and where the encoder is mounted the system will oscillate. If you "use PID's in your day job" you should be familiar with an under-damped condition.

With bowden extruders the retraction plays a large role getting good prints. The quasy-steady state is not the case. And it needs high acceleration. perhaps it's feasible when completely ignoring the encoder at a retraction (g10/g11 in Marlin)

In my opinion it's complete bullshit to add an encoder to the filament. What do you do when the extruder simple is unable to push the filament through the hole? the closed loop compensates for that, no matter what x/y coordinate the machine is having, so it might be extruding at an XY position that really didn't need the extrusion. Closed loop is one, but an overall good working printer is better IMO

Instead of dealing with encoders and PID/servo jitter problems we could rewrite the firmware to deal with the backlash.
Once we've found the right backlash value, it should be easy to use a Bowden setup with open loop, but without oozing. ( which was the reason for adding an encoder IIRC? )

Quote
to3dornottobe
In my opinion it's complete bullshit to add an encoder to the filament.

You are of course entitled to your opinion. That opinion allows for no innovation whatsoever in that area. Fine, pursue your own course!

However if someone wants to use a small DC motor they will need a closed loop.

There is no reason that a closed loop could not be done with the encoder at the hot end, and with any kind of motor, provided a better control mechanism is used other than the traditional PID control. Using a fuzzy PID control would give much greater flexibility, and is as efficient as PID. See [www.mstarlabs.com]

There is no reason a closed loop could not also be used with a stepper - it's already been done.

I have already listed some of the benefits that could be had with a closed loop, I won't bother listing them again.

---

My printer: Raptosaur - Large Format Delta - [www.paulwanamaker.wordpress.com]
Can you answer questions about Calibration, Printing issues, Mechanics? Write it up and improve the Wiki!

I meant closed loop for a bowden-fed setup. Not closed loop with filament perse. I'm aware that part didn't came through in my statement..

Did you read the rest of my post? There are some viable facts in that post that are not a mere opinion.

I'm personally a fan of direct extruders over bowden, so that was my reason for making a lightweight extruder. I know the combination of steppers and encoders are possible, If a lightweight dc motor with reduction/encoder is the way to make a powerfull lightweight direct driven extruder than I would go that route. (still, direct drive) A bowden with an encoder on the end is still worse with flexible filament than a proper direct drive (with or without closed loop) So, the bowden fed setup, encoded or not, won't work for me; i want to use flexible filament, bowden can't do that. rejected. Nothing to do with 'lack of out of the box thinking'

I see you have a delta. I'm working on a delta too, next to a core xy.. Too little time, way too little time..the extruder I'm working on is going to be used on both (lightweight, direct drive). I see you are using magnets for joints. My design doesn't use magnets but has zero backlash without the need of a steel ball and a magnet (saving weight) it's also impossible to pop loose. If your interested, just let me know, it's offtopic here. It doesn't use traditional rod-ends.

Quote
to3dornottobe
Did you read the rest of my post? There are some viable facts in that post that are not a mere opinion.

Yes of course I've read them, all of them - this is where an entirely different perspective shows up. I do agree that a traditional servo loop will not work with a decoupled encoder, and that servo loops are not needed for everything. That's fine, I'm not proposing a traditional servo loop.

I respectfully disagree that it's possible to say with any certainty that a decoupled encoder method won't work with flexible filament - it might work better as it could also account for deformation of the filament at the drive gear, at different feed rates (pressures). That doesn't mean that flexible filament might still not work in a particular bowden setup for other reasons - like increased friction.

Quote
to3dornottobe
What do you do when the extruder simple is unable to push the filament through the hole? the closed loop compensates for that, no matter what x/y coordinate the machine is having, so it might be extruding at an XY position that really didn't need the extrusion.

I think it is important to have a good understanding of the range of mm3/second that can be extruded at various temperatures with one's setup. There have been attempts to auto limit extrusion rate (Max Volumetric Speed (mm3/sec) in Slic3r), however one still needs to do torture tests to determine how much the extruder can actually deliver at a given temperature. I have such a test in the Guided Calibration suite I'm developing. For instance I know my printer will reliably extrude at 18mm3/sec for bridging/sparse infill/support, and 12mm3/sec for perimeters/solid. When you do such a test, you also discover the dropoff in extrusion that occurs - due to higher pressure on the filament causing more deformation where the gear teeth contact the filament - causing a flattening of the extrusion rate - it's not linear near the top end, its a curve. So in this case a proper closed loop would compensate for the variation in extrusion at different pressures. It would also compensate for the different deformation rates of harder and softer filaments, and more/less springy filaments. It could be self tuning.

Without knowing the max mm3 number one could still enter the wrong value in Slic3r and that would cause a stall. (I'm not a fan of auto settings at all though.)

I hope your direct drive extruder works well. If it doesn't that still something to learn from.

I'm quite interested in your joint method (in another thread). I do use tension springs also, as I learned the magnets were not enough - when it hits a curled edge etc.

---

My printer: Raptosaur - Large Format Delta - [www.paulwanamaker.wordpress.com]
Can you answer questions about Calibration, Printing issues, Mechanics? Write it up and improve the Wiki!

While this thing probably isnt 100g, I thought I'd post it here as we knife edge array(caterpillar's ass) was mentioned earlier.

http://www.3ders.org//articles/20160408-3d-cocooner-spins-nature-inspired-fiberglass-stuctures-in-mid-air.html

Edited 1 time(s). Last edit at 04/10/2016 07:29PM by MechaBits.

So did anyone experiment with their tiny motors? What are the results? Could you share extruder design? I happen to have some tiny motors lying around, but I suppose they're too tiny to work as any decent extruder.
One particular motor I could find a spec for is KP35FM2. Rated for 24V @ 500mA, it has holding torque of 700 g-cm (0.067 Nm). According to DC42's formula and my hobbed bolt diameter it could only do any good at 31.2A which would probably instantly evaporate it anyways. What are your motor specs, guys?..

Quote
toxuin
So did anyone experiment with their tiny motors? What are the results? Could you share extruder design? I happen to have some tiny motors lying around, but I suppose they're too tiny to work as any decent extruder.
One particular motor I could find a spec for is KP35FM2. Rated for 24V @ 500mA, it has holding torque of 700 g-cm (0.067 Nm). According to DC42's formula and my hobbed bolt diameter it could only do any good at 31.2A which would probably instantly evaporate it anyways. What are your motor specs, guys?..

That's exactly what I intended to add to the thread. I don't know how this thread completely dissapeared from my Followed Topics. There has been some actual concrete discussion some pages back about these tiny Nema14&co motors with gears that should have brought the weight substantially without breaking the bank. I'd like to see some reports on those experiments.

Regarding bowden, I still believe it's a good way to reduce weight and can definitely be helpful in different situations. I am not a bowden-only adept and my intention is to have a bowden and a direct extruder on the same carriage (bowden for support material, large nozzle infills), direct extruder for details, flexible filaments, that kind of stuff.

Regarding reducing weight, i've seen lately quite a trend to have split printhead solutions when each separate printhead parks itself after doing it's job. This of course requires support in the firmware but with some clever programming, there shouldn't be too much wasted time (one printhead can come to coordinates while the other retracts/etc). If there could be an open loop of singular printheads that the firmware knows how to put in the correct order for the next move there could be an efficient flow of different printheads coming and leaving the active portion of the X axis while building up the model. Even heads with pick and place capabilities could be inserted in the row and they could fill intended gaps with electronics/etc. It should be somewhat of an inverted Hyproline.

---

RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff

Just came across this thread. I too aim for the lightest head although on a Prusa, as far as speed is concerned, the bed is a limiting factor too.
I got rid of the clumsy, heavy, slippy MK8 and built a direct drive, 30g plus 240g NEMA17, plus hot end (the only thing left with the stepper). So not much fat left to trim.

I am not convinced by the Bowden approach as a general solution unless the tube is really short, not easy with a Prusa (not the best 3D printer design but very usable for its price and ease of set up, thus my choice as a first toy).

Stepper are certainly not the best, heavy, lack of torque ... but they are cheap, simple and yes, does the job. Has anyone tried (I guess yes, but didn't find any ref.) to use drive shaft, tripod joint, universal joint, spline axle, belt and pulleys .. or even flexible drill shaft to transmit the power from a fixed stepper to the moving head ?

Edited 2 time(s). Last edit at 04/11/2016 12:04PM by MKSA.

OK, I saw that the idea of using a flexible shaft like used for drill has already be done with the Flexidrive !
Definitively better than a Bowden system.

But is there a future for FDM, it is so god damn slow.

Edited 1 time(s). Last edit at 04/15/2016 07:21AM by MKSA.

Quote
MKSA
OK, I saw that the idea of using a flexible shaft like used for drill has already be done with the Flexidrive !
Definitively better than a Bowden system.

But is there a future for FDM, it is so god damn slow.

Slow is relative. I was looking at the frame for an extruder I printed last night and wondering how long it would take me to mill it out of a block of aluminum, assuming I could do it at all. My guess is 10 to 15 hours, with plenty of room for mistakes anywhere along the way...

FDM is slow compared to SLA, but the costs are much lower, especially for larger print sizes, so I think there's still several years left in the technology yet.

I gave the cycloidal gears a chance: Using an existing NEMA8 (4Ncm) with a 3D-printed 36:1 gearbox works quite well - extrusion rates and retracts are fine: Cyclo Extruder for a delta printer

Although being created on a well tuned and precise machine, I am probably lacking precision of the 3D-printed excenter, gears and the assembly: There is some waviness in surface appearance resulting from non-uniform extrusion - my old style bowden setup shows better quality.

Lets see, what we can do here...

Looks good. Looking forward to hear anything that you might do to improve on your prints.

---

RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff

Quote
powdermetal
I gave the cycloidal gears a chance: Using an existing NEMA8 (4Ncm) with a 3D-printed 36:1 gearbox works quite well - extrusion rates and retracts are fine: Cyclo Extruder for a delta printer

Although being created on a well tuned and precise machine, I am probably lacking precision of the 3D-printed excenter, gears and the assembly: There is some waviness in surface appearance resulting from non-uniform extrusion - my old style bowden setup shows better quality.

Lets see, what we can do here...

I am looking at a smaller stepper plus gearbox too as in fact the NEMA17 has much more power than needed and it would be simpler then a Brushless DC plus servo. Although you won't beat a DC motor power to weight ratio !

Edited 1 time(s). Last edit at 04/18/2016 04:06AM by MKSA.

Powermetal, are your files available?
I'd like to rework it for the NEMA14 two-bolt stepper (50g only - 14HR08-0654S). With your solution we could bring the complete cold/hot end down to 100g combined with the new DeltaMini hot end.

Edited 1 time(s). Last edit at 04/18/2016 11:15AM by jonnycowboy.

Quote
jonnycowboy
Powermetal, are your files available?
I'd like to rework it for the NEMA14 two-bolt stepper (50g only - 14HR08-0654S). With your solution we could bring the complete cold/hot end down to 100g combined with the new DeltaMini hot end.

I am planning to put the files to a wiki page here - just in case you want to play with them. This will include parts in STEP-format and simplified parts (no blends/chamfers/...) for easier modification.

The 14HR08-0654S as proposed by Chri and others seems to be a nice lightweight and effective motor solution for such kind of application - maybe another gearbox with a modified ratio (something like 15:1) with this stepper will make it's way...

Regarding precision: Is there anybody, who has access to a precise lasercutter or watercutter, who could cut properly toleranced gears from plastics to see, whether this concept is viable at all?

I can get the parts cut at Seeedstudio once the idea is built up with 3D printed prototypes for the NEMA14 / lower gear ratio solution.

Quote
powdermetal
Regarding precision: Is there anybody, who has access to a precise lasercutter or watercutter, who could cut properly toleranced gears from plastics to see, whether this concept is viable at all?

... I can lasercut dark or NIR-absorbing materials (blue 6W-laserdiode or 60Watts fiberlaser), but precise gears are mostly made by EDM ...

---

Viktor
--------
Aufruf zum Projekt "Müll-freie Meere" - [reprap.org] -- Deutsche Facebook-Gruppe - [www.facebook.com]

Call for the project "garbage-free seas" - [reprap.org]

Tha above mentioned motor is 130g according to spec sheet?

Edited 1 time(s). Last edit at 04/18/2016 07:01PM by Mutley3D.

Quote
Mutley3D
Tha above mentioned motor is 130g according to spec sheet?

I've googled 4 datasheets:
one says 90gr, two say 100gr and another says 130gr.

The only 14HR stepper which is close to 50gr is the 7Ncm 14HR05....
But other sites say 80gr for the 7Ncm and the 4Ncm weights 60gr?!

Edited 1 time(s). Last edit at 04/19/2016 02:43AM by o_lampe.