Best low-price 3D printer?

[Corporate666]

I am thinking about buying a 3D printer that I would use for rapid prototyping small parts prior to CNC machining, and also perhaps for making one-off enclosures for little projects and such.

There are so many out there now, and I have no idea where to start.  Can someone give some suggestions on what they have and if they like it?  This is something I would probably only use every couple or few months, so I don't want something that needs constant maintenance to work reliably.  I'd like to spend up to maybe $1,000 or so (is that feasible?).

[AmirTheGreat]

http://portabee3dprinter.com/shop/

I have heard good things about this, 500 bucks will get you the kit and for an extra 200 you get it built.

[TerraHertz]

There's some discussion here:

http://blog.cubehero.com/2013/04/12/what-you-need-to-know-about-buying-a-3d-printer/
https://news.ycombinator.com/item?id=5540002

[GeoffS]

Take a look at the Ord Hadron
Designed by the guy that created Makerslide.

You can get a kit on eBay for about $500 plus whatever electronics you select, single or duel extruder etc.

[MacAttak]

QU-BD (http://www.qu-bd.com/) has some very aggressively priced machines that are coming on to the market right now - the first units are being shipped out within the next few days/weeks. They have two extrusion printer models under $1,000 which are very solidly built - all precision milled parts, no lasercut plywood or printed components. QU-BD is claiming extremely fast operating speeds with this design, capable of running in excess of 500 mm/s (they actually found that it runs so fast that the USB/PC interface was usually not keeping up, so SD card is pretty much mandatory).

They also offer a printer+CNC combo (the RPM), which I think is unique in the market. But it also costs a bit more - $1,700. Not quite as fast as the Revolution and Revolution XL printers, due to the heavier gantry needed for milling, but still probably faster than most other printers on the market today.

I have one of the smaller Revolutions on order (am in the "beta" group), so I will know first-hand what the quality is like. My hopes are pretty high though - the example prints that have been shown look extremely good.

[free_electron]

3dsystems ?

ALL the hobby machines are CRAP. you need to babysit them , parts peel off ,curl up , fall apart ... they are not turnkey machines.

the real machines are. put in a cartridge of material hit start and you will get a perfect part. every time.

This is something I would probably only use every couple or few months, so I don't want something that needs constant maintenance to work reliably.  I'd like to spend up to maybe $1,000 or so (is that feasible?).

No. you need to babysit these things and futze around with them. A no-touch machine begins at an 30k$ ...

Also, the printed objects are not a solid as the real deal. they are good for mockups at best.

[nctnico]

There are so many out there now, and I have no idea where to start.  Can someone give some suggestions on what they have and if they like it?  This is something I would probably only use every couple or few months, so I don't want something that needs constant maintenance to work reliably.  I'd like to spend up to maybe $1,000 or so (is that feasible?).

There are many out there and quality is improving. IMHO you should be willing to spend around $2000 for something reasonable but I expect prices to drop fast. I've closely inspected an example of what a cheap 3D printer can do but that wasn't very smooth. Many websites feature all kinds of objects printed by their 3D printer but few show close-ups of the results. Therefore I think the best way to see whether a 3D printer is right for you is to see it in action and inspect the objects it creates yourself.

[Corporate666]

3dsystems ?

ALL the hobby machines are CRAP. you need to babysit them , parts peel off ,curl up , fall apart ... they are not turnkey machines.

the real machines are. put in a cartridge of material hit start and you will get a perfect part. every time.

This is something I would probably only use every couple or few months, so I don't want something that needs constant maintenance to work reliably.  I'd like to spend up to maybe $1,000 or so (is that feasible?).

No. you need to babysit these things and futze around with them. A no-touch machine begins at an 30k$ ...

Also, the printed objects are not a solid as the real deal. they are good for mockups at best.

Thank you for this feedback - I think you understand what I am looking for.  I have a bunch of CNC milling and lathe machines in the shop right now.  As professional machines, they "just work".  I've tried to save money on stuff before like pick and places or CNC machines, and the babysitting and compromises required are not acceptable in a business environment where time is money.  I was hoping there was a 'real' 3D printer available for a thousand or two, but it seems like that's not the case.

I'm keeping an eye on Form Labs new SLA unit, but I guess my pipe dream of a business quality printer in a low price range will remain a pipe dream for now.

[free_electron]

You are far better off outsourcing this kind of work. Goengineer for example. Or protolabs
They have the real 3d printers.

All thos plywood and plastic machines are junk. They are good to make litlle saltshakers and figurines. Try to make anything with some reasonable tolerance requirements and it won't work.
The techshop here has three different ones. The old makerbot, the new,all metal makerbot , and an UP!.
None can reliably make two identical parts. Its enough they skip a detent and all is off.

The biiiig problem of all these low end machines is that they have absolutely zero feedback from the gantry and ar table. They run blind , assuming that, if you send 5 pulses to the steppermotor, it takes 5 steps. Well guess what... Even with a 1 part per million error rate , on a large object you may go over a million steps of your stepper motors.. You will have a glitch somewher.

This is something professional equipment dosn't do. Those machines, like you cnc's and lathes, have positional feedback. Heck even a 50$ inkjet printer uses positional feedback. There is a paperfeed encoder and a linear encoder for head position. So why don't these plastic-squirters don't use that ?
The answer is simple. Because they are all run by underpowered crapduino based hardware that couldn't cope with the flurry of information that needs to be accounted for. It takes a' brain' with a substantial amount of processing power to deal with that.

A real motion co troller is a complex piece of hardware and software.

[ecat]

You are far better off outsourcing this kind of work. Goengineer for example. Or protolabs
They have the real 3d printers.

All thos plywood and plastic machines are junk. They are good to make litlle saltshakers and figurines. Try to make anything with some reasonable tolerance requirements and it won't work.
The techshop here has three different ones. The old makerbot, the new,all metal makerbot , and an UP!.
None can reliably make two identical parts. Its enough they skip a detent and all is off.

The biiiig problem of all these low end machines is that they have absolutely zero feedback from the gantry and ar table. They run blind , assuming that, if you send 5 pulses to the steppermotor, it takes 5 steps. Well guess what... Even with a 1 part per million error rate , on a large object you may go over a million steps of your stepper motors.. You will have a glitch somewher.

This is something professional equipment dosn't do. Those machines, like you cnc's and lathes, have positional feedback. Heck even a 50$ inkjet printer uses positional feedback. There is a paperfeed encoder and a linear encoder for head position. So why don't these plastic-squirters don't use that ?
The answer is simple. Because they are all run by underpowered crapduino based hardware that couldn't cope with the flurry of information that needs to be accounted for. It takes a' brain' with a substantial amount of processing power to deal with that.

A real motion co troller is a complex piece of hardware and software.

Interesting. I've run my cnc milling machine for hours at a time and I have never experienced a stepper missing a step and as you correctly point out one step lost is obvious over the rest of the run. I have most definitely seen errors creep in due to wear on the screw threads - obvious when you feel the play, and wear in the linear guides - again obvious.

Certainly real cnc machines use servo motors, they are much more nimble than steppers and can achieve greater speeds, but even they are controlled by rotational and not absolute positional feedback?

Do you have any examples of absolute positional feedback on machines with 3 or more axes of motion, a two axis lathe sounds trivial? I'd like to see how how they do it

[Corporate666]

You are far better off outsourcing this kind of work. Goengineer for example. Or protolabs
They have the real 3d printers.

All thos plywood and plastic machines are junk. They are good to make litlle saltshakers and figurines. Try to make anything with some reasonable tolerance requirements and it won't work.
The techshop here has three different ones. The old makerbot, the new,all metal makerbot , and an UP!.
None can reliably make two identical parts. Its enough they skip a detent and all is off.

The biiiig problem of all these low end machines is that they have absolutely zero feedback from the gantry and ar table. They run blind , assuming that, if you send 5 pulses to the steppermotor, it takes 5 steps. Well guess what... Even with a 1 part per million error rate , on a large object you may go over a million steps of your stepper motors.. You will have a glitch somewher.

This is something professional equipment dosn't do. Those machines, like you cnc's and lathes, have positional feedback. Heck even a 50$ inkjet printer uses positional feedback. There is a paperfeed encoder and a linear encoder for head position. So why don't these plastic-squirters don't use that ?
The answer is simple. Because they are all run by underpowered crapduino based hardware that couldn't cope with the flurry of information that needs to be accounted for. It takes a' brain' with a substantial amount of processing power to deal with that.

A real motion co troller is a complex piece of hardware and software.

Interesting. I've run my cnc milling machine for hours at a time and I have never experienced a stepper missing a step and as you correctly point out one step lost is obvious over the rest of the run. I have most definitely seen errors creep in due to wear on the screw threads - obvious when you feel the play, and wear in the linear guides - again obvious.

Certainly real cnc machines use servo motors, they are much more nimble than steppers and can achieve greater speeds, but even they are controlled by rotational and not absolute positional feedback?

Do you have any examples of absolute positional feedback on machines with 3 or more axes of motion, a two axis lathe sounds trivial? I'd like to see how how they do it

On my CNC machines, the servos are connected directly to the ball screws with an optical encoder on the back of the servo shaft.  It is not absolute position feedback, but there is not really much chance to "lose a step".  Since the motor is mechanically connected to the ballscrew directly (no belt or gears), and the ballscrew is essentially zero backlash, if the motor (and encoder) turn, it's pretty much a sure thing that the position changed.

Most NC controls are capable of taking absolute feedback using linear glass scales.  It's an option on most lower end machines and sometimes standard on high-end machines made for moldmaking.

I think the problem with the hobby 3D printers is that they are using low powered motors and generally cheap components.  I was a bit surprised to see thin wood used as a base for a 3D motion system... all of the flexing of components and tolerances add up to create errors.  That's part of what I liked about the Form Labs SLA machine - since it uses laser scanning,if the laser is fixed, there are less moving axes to go wrong.  I would not have minded putting up with 0.010" worth of error, but many of the specs they quote seem like total bullshit.

[poorchava]

I think that failures often come from to feeble construction. Linear guides have to be PERFECTLY parallel, otherwise the carriage will jam in this place or another. And here you have a missed step.

On top of that, in screw-driven applications where I have like 200steps/rev and 4mm pitch trapezoidal leadscrew, this gives me like 50 steps per milimeter without microstepping (and about 800/mm with microstepping). With 1:1 belt drive you get faster movement, but MUCH worse accuracy.
Say you have a 20mm cog wheel on a stepper. That gives you ~26 steps/mm. Now a single missed step is a big problem.

At 800steps/mm I can live with one or two missed steps.

[amyk]

This is something professional equipment dosn't do. Those machines, like you cnc's and lathes, have positional feedback. Heck even a 50$ inkjet printer uses positional feedback. There is a paperfeed encoder and a linear encoder for head position. So why don't these plastic-squirters don't use that ?
The answer is simple. Because they are all run by underpowered crapduino based hardware that couldn't cope with the flurry of information that needs to be accounted for. It takes a' brain' with a substantial amount of processing power to deal with that.

Cheap ARM boards could be the solution... although I've an inkjet printer with encoders and it uses a Renesas H8S which I don't think is particularly powerful. I think the issue has more to do with software (firmware).

[free_electron]

You'd be amazed how much horsepower an h8 has..
But, it probably also has some motion control chip.

St has really nice stepper motor comtrollers. You can set acceleration , decelration , torque curves , and much more. Big whoppin tqfp with 80 pins.
And ot has i terface for a quadrature encoder so it will close the loop.

Aal the professional machinery, pick place, lathes , cnc, robots, waferhandlers , uses feedback from encoders. There is a reason for that.

Encoders don't really cost much.