small printer, runs off of 5v, fits in a backpack.

OK so I've made a few more enquiries and that sub $5 figure is going to be a little harder to achieve than I first thought. If sourcing over 100 units at a time l should be able to get it to about $6-$7 per unit. If sourcing >1000 units then $5 might be possible.

The NEMA17 sized motors seem to be in more demand than the NEMA14 size and the 1.8 degree type again more prevalent than the 0.9degree. Both factors I believe are affecting the pricing i.e. the more they sell of a particular type of motor, the more likely they are to offer it at a cheaper price as their per unit overheads drop dramatically.

I've been approaching sourcing activities as I do professionally where I usually source parts in the >100k range so my estimation for this project was a little off. I have no doubt the likes of makerbot, ultimaker and printrbot are sourcing enough volume to be buying sub $5 motors (probably more like $2-$4).

I'm still waiting for some replies so we'll see.

Not sure about the 7.5 deg motors with no external gearing or micro-stepping, resolution just wont be there. Feel free to run some tests and prove me wrong though ;-)

Edited 1 time(s). Last edit at 05/28/2013 02:39PM by HaniC.

HaniC Wrote:
-------------------------------------------------------
> The NEMA17 sized motors seem to be in more demand
> than the NEMA14 size and the 1.8 degree type again
> more prevalent than the 0.9degree. Both factors I
> believe are affecting the pricing.

Are we trying to get smaller motors or cheaper ones? If the target is price then we should be targeting NEMA17, 1.8 degrees motors instead.

> Not sure about the 7.5 deg motors with no external
> gearing or micro-stepping, resolution just wont be
> there. Feel free to run some tests and prove me
> wrong though ;-)

Well, let's say we try 2mm diameter rods to spin the fishing line around it. That would mean a cirumference of pi * 2mm = 6.28318530717959mm. 6.28318530717959 divided by 48 steps equals 0.13089969389957mm per step. Is that not enough for an entry-level printer?

I agree I think that will be enough for an entry level printer, it doesnt need to have the high precision that most repraps and other 3d printers have. infact I doubt anyone would notice the difference if they turned their micro stepping down to just half stepping, well at least this has been the case in my experience. now the only problem is finding a cheap source for these.

I forgot about half-stepping, so that would mean 0.13089969389957mm divided by 2 equals 0.06544984694979mm per step.

Half-stepping means we could use unipolar motors and ultra-cheap stepper drivers. Worst case scenario we need to use full step drive to get more torque and are back at 0.13089969389957mm per step.

Crazy idea but… rack and pinion? I've made some tests with a toothed gear and simply pressing the gear against a ruler with a cork backing and it moved fast enough with plenty of torque, at least in my opinion. It can move the ruler with two NEMA17 motors on it, probably more than enough weight for a first test, the stepping rate could be increased once the printer is built. It's more than decent considering the cost of the motor and driver. We should also try to use this motor for the extruder, it would keep the price low, it would make the weight to move around a lot lower and we could drive the motor faster before stalling.

If we could remove the mechanical backlash from the gearing inside the 28BYJ-48, it would work. We would have at least half-decent speeds and with the internal gearing we probably already have enough precision as long as the gear is kept at a small diameter. What about removing the backlash via software? Mach3 can do it for CNC machines, I don't see why it would work for a 3D printer, there's nothing pushing against a milling bit while cutting here.

There's always the problem of motor longevity and with plastic gearing that's a huge concern. Given the cost of these small steppers, I think I'll just make a test platform to see how long it takes to kill one.

It would also be interesting to order from different sellers on eBay, first of all there's the 5V and 12V versions but there also seems to be different manufacturers, the motors look similar but there's differences such as the mounting holes, the angle of the tabs for those holes, etc. Maybe there's one company which makes a good one for 4$ with exactly 4096 steps and metal gears and the others ripped off the design and selling it for 3$ with weird ratios and plastic gears. I've read comments on forums that some people got motors with precise gear ratios and others like me who got completely insane ones.

Another difference will probably be that we should be able to drive the 12V motors a lot faster than the 5V ones. Then again the small ULN2003 driver boards aren't included with the 12V ones but I see one listing at the moment with an L298N driver board. But the auction is for 7.34$CAD, much higher than the 5V version which is around 2.75$ with the driver board. Maybe used the 12V model for the extruder, if necessary?

Last edit to this post, hopefully…

Since the target is a small printer and (it seems) a low-cost one too, how about using smaller diameter rods? If we use the 28BYJ-48 motors, the heaviest thing would probably be the rods. Since it's smaller, the print volume will be smaller. And since the 28BYJ-48 are so much lighter than a regular NEMA17, maybe 6mm rods would be more than enough. A lot of 10 LM6UU's seems to be around 8-10$ on eBay as of this writing.

aduy, what is the target printing volume you were thinking about? Would 100x100x100mm be too small?

Probably on my 10th edit so far… I think a neat way of keeping the printer compact and portable would be to integrate the filament spool into the printer itself.
the-buccaneer puts it in a "cartridge" on top of the printer. Since I currently don't own a 3D printer, I have no idea if a spool of 1.75mm filament can be as small as 12~15cm in diameter.

One last idea… it would be nice if this model could be done with either a 3D printer, small desktop CNC machine or laser/waterjet cutting machine. I have an old Roland MDX-3, but the "new" MDX-15 has the same cutting area. I'll try to design something that only needs flat panels, 150x90mm maximum.

Also, being able to print the panels on a printer, glue the page on a sheet of material and cut by hand would be an option.

HaniC, what price do you think you could get for these motors?

Edited 9 time(s). Last edit at 05/30/2013 09:49PM by Yvan256.

ok you can compensate for backlash in the firmware, im still reading the whole thing i will edit this post later

edit: those motors, we should use those, with a small diameter pulley and thin thin fishing line they will be perfect and they are strong. if need be we can print a planetary gearbox for them or perhaps just print a gear reduction.

with my new big aluminum extrusion printer it has a very efficient build area to total size, 580x660x600 total size, 470^3 build envelope.

it uses the corexy which is very simple to implement with fishing line and the bed just moves up and down. I think we can model this little printer off of my big one, but it should be tantilus style so print almost all the parts. perhaps printed t slot would be nice.

also if we can get some tiny lead screw for cheap we can print the nuts and have a spring between two of them, then just directly hook those steppers up to the lead screw.

Edited 2 time(s). Last edit at 05/31/2013 12:22AM by aduy.

ok so if you wanna be cheap, use m6 threaded rod instead of leadscrew, you get 48 steps per mm exactly and a hell of a lot of torque.

First off my aim and motivation behind working on this is to produce a small yet functional (i.e. not just a novelty) printer that is also dirt cheap. Design goals of small and cheap are not mutually exclusive and are both equally important.

Precision

I'm currently running with 100 steps/mm for both x and y on my big printer so that gives a solid resolution of 0.01 mm. I think that an entry level printer will need to have at least a resolution of 0.1 mm to be actually usable. You're right about running a test though and this is on my TODO list - To turn down the micro stepping to half or even full on my Nema17 1.8deg motors, I'll be able to get an idea about what print quality to expect with a resolution of 0.08mm or 0.16mm.

28BYJ-48

I haven't totally given up on these just yet. As I said I've got a few more coming in the mail. I've got 5V, 12V and 1/16 geared (instead of 1/64) versions for test. I think the most expensive one was about $3.50. Will repeat the tests on my set-up once I get them.

Unipolar driver

I'm not too concerned about the cost of these. A ULN2803a IC which can drive two motors will cost about $0.30 - $0.60 depending on how you buy them. Can even just use cheap general purpose transistors (plus flyback diode) but there will be a few of them for 4 motors so maybe not worth it.

Fishing Line

This is definately the way to go to minimize cost and maximize resolution. I'm thinking a pulley diameter of 5-8mm will do. 2mm is just too small and would not withstand the force required and would break/warp too easily.

6mm rods

This is also the way I was thinking. The mass of 6mm smooth rods is about 56% that of the 8mm rods so that will definately reduce torque requirements, increase speeds etc and since the travel length for all axis' will be relatively small then 6mm should be fine.

Build Volume
My build volume goal is 120mm x 100mm x 100mm. I think this is good enough for a small entry level printer. To get decent print volume and small printer dimensions all the parts need to be also small hence maximum motor size (if using hybrid) can be no bigger than NEMA14.

The only thing that I haven't figure out is a cheap decent hot-end. Any ideas?

Also FYI some guy on YouTube has experimented with building a small linear stage using cd/dvd rom parts. This concept provides some food for thought...
Link1
Link2
Link3

ive got the hot end covered, its called the pluto. I actually asked the guy that made my hot end if he wanted to make a small bowden one a while back and it will be ready at some point in the near future. its 1.75mm only.

and yes i agree around 10 steps per mm would be sufficient. also because it is small it does not need to be super fast and infact if its too fast you get bad small parts.

and yes 6mm works, the little delta i have a picture of used some 1/4" steel rod on drilled abs plastic carriages.

Edited 1 time(s). Last edit at 05/31/2013 12:57PM by aduy.

HaniC Wrote:
> Build Volume
> My build volume goal is 120mm x 100mm x 100mm.
>
> Also FYI some guy on YouTube has experimented with
> building a small linear stage using cd/dvd rom
> parts. This concept provides some food for
> thought...

Searching for "Xbox worm motor" (or "PS2 worm motor", etc) gives a lot of low-cost results, there's some people selling more complete assemblies similar to what Ezra Zygmuntowicz is showing on YouTube but it triples/quadruples the price. I think starting with the worm motor alone and building our own assembly should yield a cheaper and more custom result.

It would mean lowering the build platform to 73x73mm, if Ezra's numbers are correct. Seeing the threads on those rods, however, makes me think the motors would require more than half-stepping which means more expensive electronics. And all the motors I've seen listed are bipolar so we can't count on using the ULN2803 ICs.

Keep us informed about the different models of 28BYJ-48, I've got some projects that could really benefit from a slow, low-cost and compact stepper motor.

Ok guys, I did a test on my current printer with half stepping giving a x/y calibration value of 12.5steps/mm.

Firstly, the noise from the X and Y steppers and also vibration increased significantly. Also as expected, at low speeds they are at their worst. This was to be expected but remember I'm using big NEMA 17 motors on this printer. Smaller motors should be more tolerable to some degree.

I also tried full stepping but I aborted that after one layer as I got too much noise and vibration to the point where I thought my printer was going to break.

I did two prints. Again both were with half stepping. The first print was Mr. Jaws and Mr Piggy in black ABS @ ~0.2mm layer height.

The first 3mm were at 20mm/s perimeter and 30mm//s infil.
The next 3mm were at 40mm/s perimeter and 60mm/s perimeter.
The next 3mm was at 60mm/s and 80mm/s perimeter.
And the last 1-2mm speed were varied as I was playing around with the speed in real time.

I took a short video during each 3mm phase of the print:

[www.youtube.com]
[www.youtube.com]
[www.youtube.com]

In the first video some of the vibration sound was caused by an endstop actuator rattling. I stopped this for the next two videos.

And pics can be found here in my album:

[tinypic.com]

The Mr Jaws print lifted a little on both ends so don't look at this part of the print to guage quality as it's irrelevant for this particular test.

My observations:

- Any Z wobble i had previously was a little accentuated probably due to extra vibrations.
- As speed increased some artifacts on the perimeters appeared similar to what backlash would look like but still ok.
- No missed steps that I could see which is good.
- Perimeters were acceptable and solid fill was pretty nice too.
- Sometimes in a couple of specific areas the infil overshot the permimeter just a touch and showed up as a vertical bump along Z.

Overall, I was really happy with the quality. However, this print did not change along the Z so I printed something else to check. I tried the octopus (also in that album).

This print was printed at 45mm/s perimeter and 60mm/s infill in blue PLA @ ~0.25 layer height.

My observations:

- Not bad.
- X skipped badly during the 2nd and/or 3rd layer and skipped mildly a few more times further up in the print. If you look at the last photo of his head you can see the print is skewed to the right which means there were several skips just about each layer. The slicer I used was Kisslicer - Problem was this can sometimes generate really small fast zigzag type paths which I believe caused most of the skipping.
- But would still be very acceptable quatlity IMO for an entry level printer.

Conclusion:
0.1mm resolution or better will be fine as I don't thing this is going to be the worst source of error. It's more likely the jumpy nature of non-microstepping will negatively contribute to the quality more than resolution. Worth persuing and running further tests IMO.

Edited 1 time(s). Last edit at 05/31/2013 10:14PM by HaniC.

see the thing that differentiates 3d printing from milling is that the filament naturally smooths it self out because the plastic comes out as a circle. I would be interested in seeing some 3d printing where the resolution is so low that it makes pixelated prints. also the fast the head travels the less noticeable the "pixelation" will be.

For now lets work on the electronics and integrating that with an arduino mega and a firmware like marlin or something that has back lash compensation. arduino megas can be had for 20 bucks now since the release of the due.

Thank you for the tests, HaniC. You proved that a printing resolution of 0.1mm isn't the end of the world. As for the vibrations from full stepping, those 28BYJ-48 are so small that I don't think it will be a problem because of their low mass.

aduy, we can further lower the price by either doing our own electronics or using another Arduino. Even at 20 bucks, it's too expensive if we can get away with using something else. An older Arduino Pro can be had for 15$ or even 10$. If grbl only requires a 3.50$ ATmega328P (and only 2.22$ for 25 or more) then I don't see why we need more expensive electronics.

There's no point in trying to save 15-20$ on the motors and drivers if we waste it somewhere else without any good reason. Cut costs on all parts, I say! :p

I'm not sure if RepRap firmwares can be flashed onto an ATmega328P though, so maybe that's one place where we cannot cut the cost, at least in the development stage. A plain Atmel microcontroller should cost less than a full-fledged Arduino.

I don't have a 3D printer but I do have an old Roland MDX-3, so if someone needs a custom-cut panel for this project, I'm available. So far I've cut 3/8" MDF, 3/8" HDF, and 1/8" to 3/8" expanded PVC. The biggest panel I can cut is 145x94mm, 3/8" deep cuts. My router bit is 1/8" diameter so keep that limit in mind for inside corners. Smallest hole I can safely drill with the machine is 4mm since it's not a plunge bit.

edit: HaniC, would you care to do a few other tests? Let's assume for a moment that the 28BYJ motors can be used, I don't think we'll ever get anywhere near the speeds you tested. My question is, how slow can we go before problems appear?

You tested at 20mm/s perimeter and 30mm/s infil, 40mm/s infill and 60mm/s perimeter and 60mm/s infill and 80mm/s perimeter. How about going backward and test 15/15, 10/10 and 5/5 but all three with 1/16 stepping? I'm asking to test with 1/16 stepping since the 28BYJ has (in theory) 4096 steps and a regular NEMA17 has 200.

Edited 5 time(s). Last edit at 06/01/2013 02:07PM by Yvan256.

Yvan256 Wrote:
-------------------------------------------------------
> Thank you for the tests, HaniC. You proved that a
> printing resolution of 0.1mm isn't the end of the
> world. As for the vibrations from full stepping,
> those 28BYJ-48 are so small that I don't think it
> will be a problem because of their low mass.

Correct! Smaller motors shouldn't be as jerky - even the nema 14s or 7.5 degree motors should be ok..

> aduy, we can further lower the price by either
> doing our own electronics or using another
> Arduino. Even at 20 bucks, it's too expensive if
> we can get away with using something else. An
> older Arduino Pro can be had for 15$ or even 10$.
> If grbl only requires a 3.50$ ATmega328P (and only
> 2.22$ for 25 or more) then I don't see why we need
> more expensive electronics.

Exactly. I don't use development kits in final projects. I don't even use arduinos. Whilst I can respect their place in the hobby market and their use by those who cannot design their own circuits, technically they shouldn't be integrated into final products as they are essentially only a development kit. The only bits we really need off the arduino are the micro, the USB->UART bridge and a few of the passives like the crystal osc. I am planning to design my own custom cost/size optimised circuit.

You're spot on about the ATmega328p. On my current printer I'm running Marlin firmware on a sanguinololu. I'm not using any LCD or SD card so the firmware compiles at around 55kbytes.The 328p only has 32kbytes flash mem and half the ram of the 644p. I haven't had a look at whether it would be possible to port Marlin down to a 328p yet but it may be if a few of the redundant features are removed and others optimised. This would require some work though. Also, according to the reprap wiki, Teacup FW can fit onto a 328p. I don't particularly wan't to reinvent the wheel regarding firmware so would probably opt for one of these options if they're feasable.

For the controlling the steppers I'm planning on using either cheap discrete logic or a small PIC/AVR/MSP430 as well as a couple of ULN2803As to handle stepper control taking in the current step + direction (+ enable/disable) inputs used. Burdening the main controller with low level stepper control will not work so this isn't an option .

> There's no point in trying to save 15-20$ on the
> motors and drivers if we waste it somewhere else
> without any good reason. Cut costs on all parts, I
> say! :p

That's the plan!! Should bring the cost of the complete electronics down to under $20 .

> I'm not sure if RepRap firmwares can be flashed
> onto an ATmega328P though, so maybe that's one
> place where we cannot cut the cost, at least in
> the development stage. A plain Atmel
> microcontroller should cost less than a
> full-fledged Arduino.

See above.

> edit: HaniC, would you care to do a few other
> tests? Let's assume for a moment that the 28BYJ
> motors can be used, I don't think we'll ever get
> anywhere near the speeds you tested. My question
> is, how slow can we go before problems appear?


Regarding the 28BYJ-28, they're not an issue with respect to smoothness at any speed. Because they are geared down, we are generally running them at higher frequencies and also because they're not as powerfull, the jerkyness in not there. Both these factors are actaully are a problem. Speed is severly limited and power/torque may just not be enough - nevermind the backlash.

> You tested at 20mm/s perimeter and 30mm/s infil,
> 40mm/s infill and 60mm/s perimeter and 60mm/s
> infill and 80mm/s perimeter. How about going
> backward and test 15/15, 10/10 and 5/5 but all
> three with 1/16 stepping? I'm asking to test with
> 1/16 stepping since the 28BYJ has (in theory) 4096
> steps and a regular NEMA17 has 200.

Ok a couple of things to mention here. The microstepping is not only increasing the resolution but allowing the whole system to run very smooth. My intention when testing half stepping of the NEMA 17s was to assess the effect on resolution. What I actually found was that it was acceptable, however the jerkyness is a little problematic. This was with the motors still at full rated power/torque.

I can reliably print at < 5 mm/s on my current printer when 1/16 microstepping so that test is not needed.

The issue as I see it is that the small geared motors are not fast/powerful enough while the bigger motors don't run smooth without microstepping (and are more expensive). We probably need something in the middle. This is where something like the 7.5degree motors come in. They might be bought for $3-4 and then we can maybe gear them down 1:2 or 1:4 externally with some printed gears to get the resolution, smoothness and torque. Not sure about backlash though.

The ideal motor would be something that had >100 steps/rev (>200 half steps/rev), >8-9 N.cm torque, high enough frequency for at least 50mm/s moves and costs less than $5.

Will keep looking.

HaniC Wrote:
> For the controlling the steppers I'm planning on
> using either cheap discrete logic or a small
> PIC/AVR/MSP430 as well as a couple of ULN2803As to
> handle stepper control taking in the current step
> + direction (+ enable/disable) inputs used.
> Burdening the main controller with low level
> stepper control will not work so this isn't an
> option .

I already have code to convert an ATtiny24 into a stepper driver, I only forgot about the enable/disable pin but otherwise the step+direction pins are converted into four coil enable pins. Since it's only running a single task in a loop I didn't bother coding interrupts or anything fancy though.

If needed, we can double the power of the ULN2803A by tying the pins together in parallel, which would mean 1A/50V for a single ICs. At under 1$ per IC I think it's a viable option.

> Regarding the 28BYJ-28, they're not an issue with
> respect to smoothness at any speed. Because they
> are geared down, we are generally running them at
> higher frequencies and also because they're not as
> powerfull, the jerkyness in not there. Both these
> factors are actaully are a problem. Speed is
> severly limited and power/torque may just not be
> enough - nevermind the backlash.

By using full step drive I found the torque to be quite respectable. As I mentioned before, I had put an MXL pulley about 15mm in diameter around the motor shaft and by pushing on the cork side of a metal ruler it was able to move two NEMA17 motors sitting on the ruler, sliding against the textured plastic of my work table.

> I can reliably print at < 5 mm/s on my current
> printer when 1/16 microstepping so that test is
> not needed.

Ok, thanks for the explanation and the confirmation of slower print speeds.

> The issue as I see it is that the small geared
> motors are not fast/powerful enough […]

See above.

> bigger motors don't run smooth without
> microstepping (and are more expensive). We
> probably need something in the middle. This is
> where something like the 7.5degree motors come in.
> They might be bought for $3-4 and then we can
> maybe gear them down 1:2 or 1:4 externally with
> some printed gears to get the resolution,
> smoothness and torque. Not sure about backlash
> though.

The Tantillus uses herringbone gears to transmit movement and from the photos I've seen it seems to print incredibly precise parts.

> The ideal motor would be something that had >100
> steps/rev (>200 half steps/rev), >8-9 N.cm torque,
> high enough frequency for at least 50mm/s moves
> and costs less than $5.
>
> Will keep looking.

Ok, keep us informed about anything you can find. I'm also impatiently awaiting your tests of those other 28BYJ motors, especially the 12V version. Hopefully it's a perfect 4096 steps per revolution.

edit: I found extra-cheap, miniature stepper motors. They're so cheap that the vendor doesn't even list any specification. Yes, that's a joke I'm not being serious. but I just found them so cheap and small that's it's ridiculous.

Edited 2 time(s). Last edit at 06/03/2013 02:55PM by Yvan256.

well if you use two in parallel maybe, they look pretty weak. The corexy can use up to 4 motors btw.

and as far as not using arduino, thats a mistake. all of the popular firmwares use an arduino. there is no reason to reinvent the wheel.

Which Arduino model is the question here. Different models use different Atmel microcontrollers. The older Arduinos used the ATmega328P which only cost around 3.50$CAD.

Edited 1 time(s). Last edit at 06/03/2013 07:42PM by Yvan256.

at least the 1280

aduy Wrote:
-------------------------------------------------------
> and as far as not using arduino, thats a mistake.
> all of the popular firmwares use an arduino. there
> is no reason to reinvent the wheel.

I completely disagree. You do need to use an arduino to still run popular firmware. I'm not sure what you think an arduino is but really it's only a microcontroller with a few other bits. All I'm saying is to ditch the official arduino platform and just use a custom board with a compatible atmel microcontroller. If we use an atmega64/128 series micro and make the same pin connections as one of the popular reprap boards then we can still run any reprap firmware as is. Better yet if we use the atmega328a and run teacup we can again save even more cost.

thats what i meant, you mentioned avr and pic early, I think we should stick with the atmel chips as most of the firmwares run on them.

aduy: Atmel has many types of microcontrollers available, AVR is simply a type of core processor, like 8051 or ARM.

HaniC: for the main Atmel IC, I'm hoping we can stay with a DIP package, so maybe go with an ATmega644P/ATmega1284P instead if possible. If the firmwares are coded properly, worst case scenario we'll only have a couple of lines to modifiy to change the pins assignments, etc. But I do agree that using an ATmega328P would save money (3.54$CAD for ATmega328P vs 9.10$CAD for the ATmega1284P)

---

I just checked the prices on Digi-Key again and the ATtiny24A really is the cheapest Atmel IC we could use for the stepper driver. We need four outputs for the motor coils (A+, A-, B+, B-) and three input lines for the control (step, direction, enable).

I keep mentioning the ATtiny24 for three reasons:
- it's cheap enough
- I have an old Atmel STK500 programmer, so I can test things on my side
- I've already made the code for it and it works fine even with no interrupts and with the internal 8MHz oscillator.

The current code is for full step drive but could easily be modified for half stepping. Depends if we use the 28BYJ (full step drive gives quite a good torque considering the motor size) or a 48 steps motor (we will need half stepping). We could even add a jumper to select between half-stepping or full step mode.

I'm not sure if it could keep up with the faster signals required for a 3D printer, but the code is pretty simple and the chip can be clocked up to 20MHz with an external crystal if needed.

Atmel ATtiny24A is 1.25$CAD (qty 1)
Toshiba ULN2803APG is 0.81$CAD (qty 1)
if we forget the other components it means 2.06$CAD to drive one motor, but for only 25 units the cost goes down to 1.42$CAD. With online services like OSHPark, we could easily target a limited run of 3 printers of this model, or as easily make a production run of 150 units or more.

Edited 3 time(s). Last edit at 06/03/2013 11:36PM by Yvan256.

aduy Wrote:
-------------------------------------------------------
> thats what i meant, you mentioned avr and pic
> early, I think we should stick with the atmel
> chips as most of the firmwares run on them.

If you read that part again carefully, I mentioned PICs and MSP430's (in addition to AVR which just meant atmel micro which arduino uses) in the use of a custom stepper driver to replace the pololu drivers. The pololu drivers are $10 each where the driver we are proposing will be around $1-$2. Main controller is still going to be an atmel controller.

Edited 1 time(s). Last edit at 06/04/2013 04:59AM by HaniC.

Yvan256 Wrote:
-------------------------------------------------------
> HaniC: for the main Atmel IC, I'm hoping we can
> stay with a DIP package, so maybe go with an
> ATmega644P/ATmega1284P instead if possible. If the
> firmwares are coded properly, worst case scenario
> we'll only have a couple of lines to modifiy to
> change the pins assignments, etc. But I do agree
> that using an ATmega328P would save money
> (3.54$CAD for ATmega328P vs 9.10$CAD for the
> ATmega1284P)

Changing pin assignments is generally a non-issue. Porting code onto a different micro is another story especially if we need to reduce code size. The Amega328 is available in a DIP package in addition to SMT packages. I just ordered this for development LINK It's cheap as.

I spent some time today looking at Teacup which can run on a 328. I managed to get it to work on my current electronics after a couple hours of tweaking. It can't run as fast as Marlin in terms of steps/s so max x/y move speeds are about 100mm/s - good enough. I'm printing a yoda as I type to test it. Will let you know how it performed. BTW it compiled at under 18kB which is pretty good. TBH I can't be stuffed porting marlin so hopefully it performs well enough.

> I just checked the prices on Digi-Key again and
> the ATtiny24A really is the cheapest Atmel IC we
> could use for the stepper driver. We need four
> outputs for the motor coils (A+, A-, B+, B-) and
> three input lines for the control (step,
> direction, enable).
>
> I keep mentioning the ATtiny24 for three reasons:
> - it's cheap enough
> - I have an old Atmel STK500 programmer, so I can
> test things on my side
> - I've already made the code for it and it works
> fine even with no interrupts and with the internal
> 8MHz oscillator.
>
> The current code is for full step drive but could
> easily be modified for half stepping. Depends if
> we use the 28BYJ (full step drive gives quite a
> good torque considering the motor size) or a 48
> steps motor (we will need half stepping). We could
> even add a jumper to select between half-stepping
> or full step mode.
>
> I'm not sure if it could keep up with the faster
> signals required for a 3D printer, but the code is
> pretty simple and the chip can be clocked up to
> 20MHz with an external crystal if needed.

It will keep up with no issues. Think of it this way:

With a maximum speed of 100mm/s and X/Y axis steps/mm of 100, our max steps/sec will be 10,000. With the processor running even as low as 1MHz we have 100 cycles to execute each step. Should be a breeze for any micro regardless of coding method used. I'm even considering 2 motors per micro or even 4 if the pin count/cost balances out. In terms of processing power alone a single micro can drive all four motors with plenty of processing power to spare.

> Atmel ATtiny24A is 1.25$CAD (qty 1)
> Toshiba ULN2803APG is 0.81$CAD (qty 1)
> if we forget the other components it means
> 2.06$CAD to drive one motor, but for only 25 units
> the cost goes down to 1.42$CAD. With online
> services like OSHPark, we could easily target a
> limited run of 3 printers of this model, or as
> easily make a production run of 150 units or more.

I use itead studio for small batch prototypes of PCB only. Can cost as low as $14 for 10PCBs shipped. For dev we can source the other parts from even ebay. Bigger runs we have several options.

HaniC Wrote:
> I spent some time today looking at Teacup which
> can run on a 328. I managed to get it to work on
> my current electronics after a couple hours of
> tweaking. It can't run as fast as Marlin in terms
> of steps/s so max x/y move speeds are about
> 100mm/s - good enough. I'm printing a yoda as I
> type to test it. Will let you know how it
> performed. BTW it compiled at under 18kB which is
> pretty good. TBH I can't be stuffed porting marlin
> so hopefully it performs well enough.

As long as it works properly, then it's a valid choice. I'm pretty sure at least one firmware will keep using the ATmega328 for at least some time because a lot of the older 3D printer controllers use it.

> With a maximum speed of 100mm/s and X/Y axis
> steps/mm of 100, our max steps/sec will be 10,000.
> With the processor running even as low as 1MHz we
> have 100 cycles to execute each step. Should be a
> breeze for any micro regardless of coding method
> used. I'm even considering 2 motors per micro or
> even 4 if the pin count/cost balances out. In
> terms of processing power alone a single micro can
> drive all four motors with plenty of processing
> power to spare.

I've checked Digi-Key again, and I found the AT89LP51 for 1.40$CAD. It's a 40-pin DIP with a 8051 CPU core but from what I've read both AVRStudio and my STK500 can handle it. It can even be clocked up to 25MHz at 5 volts. If it's fast enough to handle everything without interrupts, then it could drive all four motors (X,Y,Z,E). It's internal oscillator can only run up to 1.8432 MHz however, so add about 40 cents to the cost for a crystal. Still, that would mean around 2.00$CAD to drive the motors, 3.25$CAD for the ATmega328P and 4.05$CAD for five ULN2803A, so I'm guessing with all the connecting hardware and PCB, the whole electronics system should fall around 15-20$CAD at the most.

So if we can either use those 28BYJ's or find 7.5 degrees motors for around 4$ each, that means around 40$ total so far. My biggest cost concern is now the extruder, the Z-axis couplers and the 6mm smooth rods. The cheapest I found on eBay means 15$ for the rods alone, putting the total at around 65$. Can we get a small extruder with hot end for 35$?

Instead of LM6UUs I suggest using brass bushings. They're about the same price on average but they're much lighter, a bit more compact and will be less noisy. On a Mantis CNC they're just glued on the plates and those machines work just fine.

> I use itead studio for small batch prototypes of
> PCB only. Can cost as low as $14 for 10PCBs
> shipped. For dev we can source the other parts
> from even ebay. Bigger runs we have several
> options.

That's a lot cheaper than OSHPark, however I've read that the quality from OSH Park is better than the other two. OSH Park is also closer to me (Canada) and their service is nearly twice as fast, so for only a few boards I'd rather use them. They also offer lower prices if you order 150 square inch or more.

Edited 1 time(s). Last edit at 06/04/2013 11:41AM by Yvan256.

' I think we should stick with the atmel chips as most of the firmwares run on them. '


I agree, for the main reason of support. In general there is a huge community even for just the arduino for support of the hardware which means there is a large knowledge base for issues.


I was evaluating at one point for development for cortex processor on arm, and determined that would be a costly approach because support would be non existent. now that the duo exists from arduino, this may change how arm processors are used, and allow the same compiler to make code. Still keeping with the arduino platform seems very wise.

It is a safer bet that hardware issues with be addressed as long as people know the hardware.

Also there are issues with the open source commitment of pic processors such as chipkit, and compatibility issues that arise from different hardware mimicking avr. for example even though chipkit can run inside of arduino ide there are different commands to access the i/o directly which is what most reprap firmwares now do.

IMHO.

@jamesdanielv

With no disrespect intended this is why I personally despise arduino.

While I love the fact that so many people are tinkering with electronics that probably wouldn't have before due to the arduino project, the hacker/maker community is now confused about what development is really about. The arduino is essentially an atmel micro with a few extras that allows one to quickly prototype one-off projects etc. The "Atmel micro" part is what is important when it comes to maintaining compatibility of firmwares not the "arduino" itself. There is a distinction and there isn't if you know what I mean. Arduino is just a name not a microcontroller.

Code that runs on arduinos doesn't need to only run on an arduino! The arduino is just a piece of hardware (just like any) that happens to use atmel variety of micro controllers and that has been widely adopted by the hobbyist community. Thats all.

Bottom line is we are not talking about changing the micro just dropping the overpriced and unnecessary arduino w.r.t. this particular project.

@Yvan

Yoda came out really well. Pics tomorrow. Some minor issues with PID temperature control and pauses during retracts but I should get to the bottom of them. Very promising indeed.

I'm pricing the electronics at a little less than what you worked them out to be. I have another really efficient hardware concept that I'll be testing out in the next couple of days. As for other bits and pieces, there are definatly cheaper sources.

Shit... Look at the time. I'm off to bed. :-)

HaniC Wrote:
> Yoda came out really well. Pics tomorrow. Some
> minor issues with PID temperature control and
> pauses during retracts but I should get to the
> bottom of them. Very promising indeed.

I can't wait to see the photos!

> Shit... Look at the time. I'm off to bed. :-)

So… I guess I'll have to wait. :-)


> I'm pricing the electronics at a little less than
> what you worked them out to be.

Well, all the costs I listed in my previous message are for single units from Digi-Key. Checking the prices for 10 to 25 units would lower the cost.

Prices are in Canadian dollars, rounded up to the nearest cent. Dirt cheap parts like resistors and capacitors are omitted for now.

ATMEGA328P (1 per printer)
1 @ 3.54
25 @ 2.22

ATS16B (16MHz crystal, since the firmwares expect an Arduino/Pro/mini and those are clocked at 16MHz AFAIK)
1 @ 0.39
10 @ 0.33
50 @ 0.29

ATTINY24A (need 4 per printer, the two Z-axis motors share the same signals and assuming the internal 8MHz oscillator is fast enough)
1 @ 1.25
25 @ 0.79
100 @ 0.70

ULN2803APG (need 5 per printer if wired in parallel to increase power to 1A per IC)
1 @ 0.81
10 @ 0.71
25 @ 0.62
100 @ 0.54

OSTTC062162 (to connect the motor wires to the board, need 5 per printer - I even tried to check prices for the 2 and 3 positions connectors, but this is the cheapest option that wasn't screwless or press-fit)
1 @ 1.04
10 @ 0.91
50 @ 0.82
100 @ 0.79

Total electronics parts cost for 25 printers, minus the resistors, capacitors, voltage regulators, other various parts and the PCB itself:
2.22, ATMEGA328P
0.33, ATS16B
2.80, ATTINY24A
2.70, ULN2803APG
3.95, OSTTC062162

Total: 12.00$CAD

Some components increase the total cost quite rapidly, like those OSTTC062162 connectors.


> I have another really efficient hardware concept
> that I'll be testing out in the next couple of days.

Care to share the concept with us? Even a photo of a sketch drawn on a napkin. The more people are thinking about it, the better!


> As for other bits and pieces, there are definatly cheaper sources.

You do seem to have better sources than the rest of us.


Just so we're all on the same track, though… have we decided yet what type of printer we're talking about (will it be similar to a Prusa, a Rostock, a FoldaRap, etc) or will that be determined by the costs?

ok first of all the point of this was to make a small portable printer that can run off 5v using these cheap motors.

second I like the arduino because i can go and get it in town or on amazon with free shipping.

third, theres no reason to reinvent the wheel, who is going to be able to go home and manufacture these cheap but incredibly complecated electronics

fourth who is going to write the code, I certainly am not up to the task.

If you want to make it cheap, lets make an actual budget, like 100 or 200 dollars.

aduy Wrote:
-------------------------------------------------------
> ok first of all the point of this was to make a
> small portable printer that can run off 5v using
> these cheap motors.

I think we all agree on that: small, portable and low-cost.

I have no idea if a hot end can run on 5V, but if we can run all the electronics and the motors on a single 5V power supply then it's a good start. I don't know about the hot end, however. And given the target build volume, I think we'll forget about a heated bed. I wouldn't mind being limited to 1.75mm PLA.

From the test I've made on my side, the 5V version of the 28BYJ motors I have, using full stepping, is more than strong enough in my opinion but has a fractional stepping ratio which could complicate things. I'm waiting for HaniC to receive the other models of 28BYJ motors to see if the 12V models are even more powerful and if the 1/16 and 12V models have a perfect 1024/4096 steps per rotation. Since all those motors cost about the same on eBay, it doesn't change much apart from knowing which vendor to buy from.

How small of a printer are we talking about, though? Is a build volume of 100x100x100mm enough? I'm guessing we can look at the Tantillus for a good model of compact, portable printer. He didn't seem to have a low price target, however his idea of using fishing line to transmit motion will sure help in lowering our cost, weight and precision. A single roll of 65lb Spectra braided line should be enough for a dozen printers or more. That's around 1$ of fishing line for the X and Y axis instead of costly belts. I'm still curious about the rack and pinion option, though.


> second I like the arduino because i can go and get
> it in town or on amazon with free shipping.
> third, theres no reason to reinvent the wheel, who
> is going to be able to go home and manufacture
> these cheap but incredibly complecated
> electronics
> fourth who is going to write the code, I certainly
> am not up to the task.

If we design a board to use the ATmega328P, it will run exactly the same firmware as an Arduino Uno, Duemilanove, Pro or mini. Because that's what these models of Arduino are: an ATmega328P microcontroller with a specific board design and pins headers. Even a "real" Arduino Pro Mini is an Arduino even if it's not compatible with the standard shields. Further example from Arduino themselves, an ATmega328P on a protoboard can be an Arduino. It's just a name and a bootloader. We won't need to code anything special as long as the firmware can run on an older Arduino Uno/Duemilanove/etc.

The thing is, a "real" Arduino board is not exactly cheap compared to the cost of its components. After all, people are trying to make a profit by selling these. But if we're trying to keep costs down on this project there's no point in paying 20$ for a 3.50$ chip. I also think we should try to design the PCB to be single-sided so people can make their own board at home. That's also why I suggested through-hole components. Even if we use an online PCB service and go with double-sided or even a 4-layers PCB, it's still easier to solder through-hole components.

By using the same standard components on our custom board, nothing will prevent anyone from running this printer with a real Arduino board and the small driver boards that come with some 28BYJ motors on eBay.

However:
- it will probably cost more than an all-in-one board
- the end result will just look less neat than a single board because of all the wires connecting the different boards together
- it will be less compact

In my opinion, the last two points are not desirable for a portable printer.

Also, not everyone can buy an Arduino locally and Amazon doesn't offer free shipping all over the world. As an example, it would probably cost me something around 40$ or more to order a 20$ Arduino from a USA vendor. Shipping costs, duty fees, brokerage fees.

> If you want to make it cheap, lets make an actual
> budget, like 100 or 200 dollars.

Just for the fun of it, I think we should target 100$ or less. It's going to get more headlines. "The first complete, 100 dollars 3D printer".

edit:
1. I just thought about something. If we go with the 28BYJ motors, we won't need the OSTTC062162 but I'll need to find the proper connector for the motors.
2. If you were worried about the part were we're talking about coding something, it's only about the stepper driver circuit to replace those "costly" Pololu stepper drivers. And I already have the code for that to run on an Atmel ATtiny24.
3. I just noticed that 16x2 LCD modules on eBay are only 2.50$, so if a 3D printer firmware compatible with the ATmega328P offers LCD+SD card too, we should choose it. The printer will be more portable if you don't have to drag a computer along with it.

Edited 12 time(s). Last edit at 06/04/2013 10:25PM by Yvan256.