Belt Versus Screw Drive

I am trying to fully understand the trade offs between these two approaches. There are so many linear motion assemblies that it gets hard when designing from scratch to determine the best one to go with.

At a high level, there are belt driven assemblies and screw driven assemblies. In some cases the belts drive screws as well.

If anyone cares to share there experience it would be useful for me. Naturally accuracy is most important to me meaning a high granularity of control and speed ranks a nearby second as you don't want to wait 3 days to print a cup.

So what is the best way to go to optimize accuracy with decent or good speed?

Kevin

I suspect the reasoning behind screw drive is familiarity with CNC equipment?

Obviously there is more to it than that, however I would think that people would tend to stick with what they know if they aren't familiar with both.

I am interested in what provides the best combination of accuracy + speed. Given all the experimentation here I would think there should be some really good experience in this area...

I built a screw drive McWire. I build it as a screw drive because I could acquire the parts from McMaster and the local hardware store and easily put it together with limited tools and experience.

If I had a choice, watching how painfully slow the McWire is, I would go with belt drives. In fact I am experimenting with converting it to belt drives.

I've managed to get my McWire up to 8mm/s by using different motors and drivers but its still much slower than a belt drive unit.

I can't speak for anyone else but for me it was simple really. To make a RepRap you need printed parts. If you don't have access to a RepRap to make printed parts then you have to make something like the McWire.

If they had a design for a RepStrap that was belt based I would do that instead but I went with what was already documented.

-- Chris

I think accuracy tends to rely more on the motor than anything else. While belt drives rely on their tension to create movement, the screw drive relies on it's rigidity. Both have a tremendous amount of accuracy, but it's a matter of how small you can make the motor move.

One way I've seen mentioned to make a screw drive faster would be to change the screw to something where the pitch on the screw is wider. With this, a single turn moves the captured nut further.

I think in the case of the McWire bot, the screw pitch (10-24) was chosen for it's accessibility. Fortunately, this is pretty easy to swap out for a wider pitch. new leadscrew and a new captive nut and you're set.

The other option people are looking at is the rack and pinion gear. This sort of marries the belt drives speed with the leadscrew's rigidity.

good stuff guys..."thanks".

Has anyone done a rack and pinion with RepRap style printer here? It would be good to compare that versus belts\screws from an accuracy\speed perspective as well before i make a final decision...

Kevin

The herringbone racks that forrest has been playing with look really lovely. By their very nature they have little backlash and are self-centering so they can contribute alignment as well as motion.

---

-----------------------------------------------
Wooden Mendel
Teacup Firmware

I've been able to run the screw drive as fast as 25 IPM on Contraptor assemblies. However, the reliable speed is about 15 IPM which is pretty slow. This is with 1/4-20TPI all-thread screws and 60 oz*in motors driven by Reprap SMD v1.2. It's difficult to push the motors past 400 RPM, and at 1/20" per revolution this is about as fast as it gets.

The belts on the other hand go a lot faster with the same motors - 2" per revolution on my setup, which gives speeds up to 600 IPM: [www.contraptor.org]

However, since the belts stretch some, they are not a good idea with higher loads (like fast acceleration and milling). Since I want to build a hybrid machine. I'm going to try multistart leadscrews with 1/4" travel per revolution - this should get the top speeds to 60-80 IPM range.

---

Vitaly
[contraptor.org]

Most of the CNC threads I have followed on cnczone.com (I think?) seem to end up with ball screw arrangements. I have not delved very deeply but they superficially anyway seem to have the best characteristics for heavier loads. Ball screws offer the accuracy of threaded drive with the speed of belts, tradeoff is that generally ball screws are costlier.

On what I am working with I am employing rack and pinion, it seems to offer a good balance for the relatively lighter loads found in plastic printing planes. I have been using the nylon rack from mcmaster, it is a little "mushy" and hard to mount. I am not waiting for the steel rack to get delivered and will give that a try.

I found belts to be a pain, you need to have pulleys, tensioners, the belt, and need to route the belt around everything else.

Having looked at contraptor in detail can you explain what benefits it has over something like 8020 (8020.net)? I really didn't get it once I started playing with 8020.

goinreverse Wrote:
-------------------------------------------------------
> Most of the CNC threads I have followed on
> cnczone.com (I think?) seem to end up with ball
> screw arrangements. I have not delved very deeply
> but they superficially anyway seem to have the
> best characteristics for heavier loads. Ball
> screws offer the accuracy of threaded drive with
> the speed of belts, tradeoff is that generally
> ball screws are costlier.

Right, the cost is one issue and then as I understand, the machine has to be really rigid to take advantage of the higher ballscrew accuracy. Another thing is that I'm not going to mount anything beyond Rotozip on the toolhead, since this is meant to be a desktop machine running Dremel. At these loads, the ballscrew would seem like an overkill.

> On what I am working with I am employing rack and
> pinion, it seems to offer a good balance for the
> relatively lighter loads found in plastic printing
> planes. I have been using the nylon rack from
> mcmaster, it is a little "mushy" and hard to
> mount. I am not waiting for the steel rack to get
> delivered and will give that a try.

Rack and pinion is a great alternative. I want to be able to use it in Contraptor eventually, but that will need some work. Why is the nylon rack hard to mount btw?

> I found belts to be a pain, you need to have
> pulleys, tensioners, the belt, and need to route
> the belt around everything else.

When I checked racks on mcmaster, they seemed more expensive comparing to belts.

> Having looked at contraptor in detail can you
> explain what benefits it has over something like
> 8020 (8020.net)? I really didn't get it once I
> started playing with 8020.

There are several considerations. While the structural 80/20 is pretty inexpensive, their linear motion components are not exactly cheap. By comparison, sliding elements for the angle are a lot cheaper.

The raw angle is cheaper than 80/20 (which is $2.75/ft) and perforated angle will be cheaper than this in volume (not yet). 80/20 weighs more. T-nuts are expensive (maybe there are cheap sources but I couldn't find them for 10-24 thread).

That said, 80/20 can be used with Contraptor instead of and together with angle, as the system is meant to be compatible with T-slot, gridbeam etc. In fact I've sketched a couple of designs using primarily 80/20 and Contraptor linear motion components.

---

Vitaly
[contraptor.org]

The nylon rack doesn't come with mounting holes and it has an inherent bend because of the molding process. It just doesn't really want to stay where you put it very well. I thought the steel was expensive until I realized it was coming in 4 or 6 ft lenghts. 6ft for $50 in steel with predrilled mounting holes I think or $6 for 1 in nylon... I think sdp-si has cheaper steel rack I think but I can never easily find the right parts on their site.

So rack $8.30 per foot, nylon gear $3 = $11.30 vs. belt $12, pulley $6 X2, tensioner $4, bearings/mounts $4 = $26

If you traversal lengths are much longer, say over 3 feet then rack may not be cost effective at all.

Thanks for the 8020 comparison. One thing on the linear motion is that I think precision rods may be a good fit to combine with 8020 cheaply as their milling/drilling options will put the mounting holes already in there. If you aren't doing anything heavy (their linear motion assemblies are rated to hundreds or thousands of pounds) you can use the raw pads or the reverse t-slot mounts which are only about $3 per pad.

For contraptor you may want to talk with 8020's "kitting department". They will build a sku that has all your parts already defined, cut/mill/drill them all, package it up all nice, and drop ship it to end users. In any volume say over 100 units there are dramatic price reductions that can be negotiated. I would also think that the cost savings of not having to actually do anything might make them competitive. I am not a salesperson for them or anything, I have just been playing with their stuff in detail and I think it is very slick.

I just quoted the contraptor structural set in the retail pricing and it comes to $395 already cut, ready to go. I bet you could negotiate that down to under $300 in any sort of volume as most of the cost is the per each milling/cut fee. Get it setup in a few days, unlimited qty, no inventory, ships to buyers in 1-2 days.

I think the linear motion could be done under $350 also

Just FYI.

>
> Having looked at contraptor in detail can you
> explain what benefits it has over something like
> 8020 (8020.net)? I really didn't get it once I
> started playing with 8020.

Thats easy "cost".

Plus the whole self replication idea is more possible with angled aluminum if you have a multi-tool machine that takes the aluminum and can cut and drill it...although this would not be my own goal.

Maybe people don't think its self replication if the supply stock is angled aluminum versus something that is printed, but to me the "inputs" could be anything thats generically available.

maitri982 Wrote:
-------------------------------------------------------
>
> Maybe people don't think its self replication if
> the supply stock is angled aluminum versus
> something that is printed, but to me the "inputs"
> could be anything thats generically available.

I think in terms of RepRap, Self Replication is considered the ability for the printer to produce itself from raw materials. If we extend the definition, the McWire becomes a true reprap as well...

SDP-SI racks:
[sdp-si.com]
Very expensive... looking at 7-12 inches...

goinreverse Wrote:
-------------------------------------------------------
> So rack $8.30 per foot, nylon gear $3 = $11.30
> vs. belt $12, pulley $6 X2, tensioner $4,
> bearings/mounts $4 = $26

I didn't count that as extra cost since by that time Contraptor already used pulleys and belts to couple the lead screws.

> If you traversal lengths are much longer, say over
> 3 feet then rack may not be cost effective at
> all.

Belts seem to work fairly well over short distances - it's the longer travels where the flex is more pronounced and rack/pinion seems to be the preferred option. Another consideration is that with the belt your moving stage can be extremely light, while with the rack you've already got the motor on it. Tradeoffs..

> Thanks for the 8020 comparison. One thing on the
> linear motion is that I think precision rods may
> be a good fit to combine with 8020 cheaply as

Oh yes they are. They should work well with angle based structures as well. At a certain length though they become too heavy/expensive to maintain accuracy, but they certainly work well over short travels. The oilite sleeve bearings are super cheap, too.


> pounds) you can use the raw pads or the reverse
> t-slot mounts which are only about $3 per pad.

I've not looked at 80/20 raw pads, how do you mount stuff to them?

> For contraptor you may want to talk with 8020's
> "kitting department". They will build a sku that
> has all your parts already defined, cut/mill/drill
> them all, package it up all nice, and drop ship it
> to end users. In any volume say over 100 units
> there are dramatic price reductions that can be
> negotiated. I would also think that the cost

Contraptor emphasis is on hobby/amateur prototyping of cartesian machines, so the idea is to have as many motion approaches in the set as possible at a price as low as possible. 80/20 does that and a lot more but on industrial scale, so when you count everything, including power requirements, it comes out, well, pricey.

However, this is sensible approach for "production" machines (including Contraptor ones) that need easy-to-assemble kits.

> savings of not having to actually do anything
> might make them competitive. I am not a

Arguably Ponoko wins that one, but it doesn't have the instant gratification aspect

> I just quoted the contraptor structural set in the
> retail pricing and it comes to $395 already cut,
> ready to go. I bet you could negotiate that down
> to under $300 in any sort of volume as most of the
> cost is the per each milling/cut fee. Get it setup
> in a few days, unlimited qty, no inventory, ships
> to buyers in 1-2 days.

That's 1010? At $300 that's roughly $6/ft, perhaps with smaller number of cuts it'll be less. 1010 in some of the Contraptor lengths might work in many cases. Why are T-nuts so expensive btw? Do you have a cheap source for them?

> I think the linear motion could be done under $350

I think it would greatly depend on axis travels, load reqs, power reqs, etc.

---

Vitaly
[contraptor.org]

I think in terms of RepRap, Self Replication is considered the ability for the printer to produce itself from raw materials. If we extend the definition, the McWire becomes a true reprap as well...

1) Self replicating subtractive fab machines are interesting.
2) A good RepStrap is a good RepStrap.

I was comparing the linear motion to the existing linear motion sub-package substituting 8020 parts where applicable.

If you can get ponoko to ship you something in less than 30-40 days godspeed to you, also their tolerances seem to be pretty wild sometimes.

The quote was 1010, I show it as about 50ft so consistent with you. The economy t-nuts are pretty cheap at $0.14 in 1 count? There is also nothing stopping anyone from using an appropriately sized square nut or even a regular hex nut. The ton of small cuts do effect the cost.

My take is between waiting XXX (days, weeks) to buy for $200 (which basically means I will never buy it, no impulse) or buying right now ready to use this weekend for $395, the buy now is a no-brainer. If you had stuff I could buy immediately you would have had my business months ago. Also realize that volume is everything, if you can demonstrate volume you have a ton of leverage on 8020 and whichever distributor you select on price. If you could move 1,000 units a year (at $300 or $400) that is more 8020 than most of their distributors (with a few notable exceptions) do in a year total across all that distributors customers. Seems like the "a bit under $1,000" price point seems to work pretty well for most MAKE type projects. Yes there are people who can't do that but they are vastly out-numbered by people who can/will, in the US anyhow. You can always have a cut-it-yourself 8020 kit (carbide mitre saw blade does a pretty good job), at least you don't have to drill all those holes...

Raw pads have a hole and can be mounted with a t-nut or the top mounts have a profile that slides into the t-slot.

Just some thoughts...

> If you can get ponoko to ship you something in
> less than 30-40 days godspeed to you, also their
> tolerances seem to be pretty wild sometimes.

I didn't know there was such high demand

> The quote was 1010, I show it as about 50ft so
> consistent with you. The economy t-nuts are pretty
> cheap at $0.14 in 1 count? There is also nothing

That's $14 for a 100 vs $1.4 for 100 hex nuts. Little by little it adds up.

> stopping anyone from using an appropriately sized
> square nut or even a regular hex nut. The ton of

I tried, they don't work as well.

> My take is between waiting XXX (days, weeks) to
> buy for $200 (which basically means I will never
> buy it, no impulse) or buying right now ready to
> use this weekend for $395, the buy now is a
> no-brainer. If you had stuff I could buy
> immediately you would have had my business months
> ago. Also realize that volume is everything, if
> you can demonstrate volume you have a ton of
> leverage on 8020 and whichever distributor you
> select on price. If you could move 1,000 units a
> year (at $300 or $400) that is more 8020 than most
> of their distributors (with a few notable
> exceptions) do in a year total across all that

Point taken and I appreciate the advice.
I'll run the numbers on mixed structural kit - shorter angle + longer T-slot.

> distributors customers. Seems like the "a bit
> under $1,000" price point seems to work pretty
> well for most MAKE type projects. Yes there are
> people who can't do that but they are vastly
> out-numbered by people who can/will, in the US

I think it's a matter of perspective. I happen to think that the magic number is somewhere between $300 and 400. I would expect a $1,000 machine to be almost turnkey. There are several desktop CNC kits just under $1,000 out there that look decent, but they haven't been getting a lot of attention (Zenbot, Fireball CNC).

---

Vitaly
[contraptor.org]

I should clarify I meant under $1000 including electornics, frame and linear motion. Good luck, I look forward to seeing what you come up with.

On a screw drive, couldn't you also add gears to increase RPM ? Say 1:2 or 1:3 or something, assuming the motor is strong enough.

OK, I have been very interested in this topic because of materials I have in my companie's "scrap pile" from some of our R&D endeavors, so I am jumping in.

I am in the process of collecting materials to build a Mendel. I have all the framework pieces, have a friend with a Makerbot who is making me some of the RP parts. In my R&D leftovers I have a commercial screw drive system consisting of a Nylon square nut and an anodized aluminum shaft. It moves about 9mm per revolution, nice and smooth, no real backlash. According to my calculations, the pitch diameter of the Mendel drive gear is about 90 mm.

This means that the drive with this screw is 10 times smaller per revolution.

My question, is the drive speed maximized in belt driven systems? in other words, what percentage of speed is used on a standard 1.8 degree stepper motor? Is it railed? 10%? I also have a pile of old stepper motors (the larger ones, not the 17 NEMA units called for in the Mendel design, so torque to drive the screw shouldn't be a problem, assuming my Mendel style stepper drivers can handle the current. In the experience of the users out there, how much of the speed capacity of the standard stepper motor is used?

In short, do you have a lot of overhead not used, or do you pretty much max out the motor?

I want to avoid belts on a design eventually. (I bough twice as many allthread and smooth rod as I need, to make one machine per the Mendel instructions, then start a second with slider bearings and screw drive system if possible)

> My question, is the drive speed maximized in belt
> driven systems? in other words, what percentage
> of speed is used on a standard 1.8 degree stepper
> motor? Is it railed? 10%? I also have a pile

I think it depends. Many steppers come with torque curve which shows how much torque you can expect at a given RPM. So in theory, your RPM will be limited by the torque required keep your stage moving.

> handle the current. In the experience of the
> users out there, how much of the speed capacity of
> the standard stepper motor is used?

With the all-thread screw drive I've been able to get my steppers up to about 500 RPM max, with stalls every now then. Reliable speed is in 300 RPM range - works out to be 15 IPM on 1/4-20 all-thread, which is too slow. My steppers don't have a datasheet, so I don't know how much speed capacity I'm getting vs rated, but I'm pretty sure it's as good as it gets. Maybe more expensive steppers+drivers+power could get that speed up, but I feel it would be an overkill for an all-thread screw. There is also whipping issues as leadscrew RPMs go up.

> I want to avoid belts on a design eventually. (I
> bough twice as many allthread and smooth rod as I
> need, to make one machine per the Mendel
> instructions, then start a second with slider
> bearings and screw drive system if possible)

Check our rack and pinion, or ACME leadscrews. Larger ACME have longer travel per revolution, there is also small diameter multistart ACME which can give you 1/4" travel/rev but they're more expensive. Rack and pinion goes even faster.

---

Vitaly
[contraptor.org]

So, with these Berg nuts and lead screws I have available to use, at 9mm, or .38" travel per revolution, at 300RPM I would get approximately 112 IPM travel speed. This is about 7 times the travel you get with a 1/4 20 threaded rod. If I have enough torque. If my whipping isn't bad. If I can get that many RPM with the Mendel stepper drivers I am using. If the fates permit.

The advantage of using these Berg lead screws/nuts versus say, rack and pinion, is I HAVE these in my office right now, no cash outlay needed. They are 12 starts, so I have a 1/4" dia lead screw with a fairly high travel. Nice, smooth, tight.

So, my question for current Mendel or Mendel-like systems, how many IPM is standard travel for x and y on usual 3d printing on this type of system? I am sure I have read it on some forum, don't remember now, I will have to search, unless someone can point me the right way.

skeinforge default settings seem to use 500 to 1000 mm/min (20-40ipm), and your top speed is only limited by layer adhesion and freezing.

---

-----------------------------------------------
Wooden Mendel
Teacup Firmware

So, theoretically, to get mid-range in that setting (750 mm/min) I would just need to get a stepper motor with sufficient torque turning at 84 rpm with my Berg leadscrews and 9mm per revolution. 111 RPM would give me 1000 mm/min. Hmmm, possible, I suppose. I just need to figure out if I can get 100 RPM from my stepper motors with standard Mendel type drivers.

> My question, is the drive speed maximized in belt driven systems?

Only a few people have pushed the belt drive to its limits. On my Darwin system, the speed is limited by the extrusion rate, with lots of head room on the xy feedrate. When I try to increase speed, I have trouble maintaining temperature.

---

---

Darwin clone, Gen 2 electronics, Arduino Duemilanove w/ AtMega328, 5D Firmware, Pinchwheel extruder
[www.codeerrors.com]

I would like to point out that there are more choices than just lead screw versus belt drive. Some have mentioned rack and pinion as being faster, but moving the motor onto the stage.

Another approach is to replace the belts with aircraft cable (braided wire cable, not sure what it is called in Europe). This has the advantage that it can bend in several directions, allowing the wire to wrap (possibly multiple times) around a pulley, then down one edge of the stage right next to the linear slider, then around a pulley at a diagonal to the far corner of the stage to another pulley, back down the farside right next to the slider, then another pulley diagonally to the first corner and back onto the motor's pulley. This means that the force applied to the sliding stage is balanced on both sides, no torquing moment, and the forces attach right next to the linear slides to further reduce torques and side loads. With sufficiently rigid frames and pulleys, this could handle moderate side loads from milling. If the wire cable is wrapped around the pulley enough times, you can actually screw it down to the pulley in the middle of the wire so that the cable never drifts or slips. That way you get the speed of a belt drive with the rigidity of the lead screw drive.

Another approach not commonly known or used is called a 'taut band'. A long narrow piece of thin, spring steel shim stick is cut with a "U" shape from end-to-end, leaving a long tongue in the middle. The tip of the tongue is bolted onto one edge of the stage, and the end that the tongue came from is bolted down at the other end. The metal is wrapped around a metal shaft/axle enough times so that as the motor rotates the shaft, it pulls one end closer while the other end unrolls, but always the bands are under tension, hence the name 'taut band'. This will have no backlash, quite a bit of strength, and depending on the diameter of the shaft/axle, it can be very fast.

Mike

> My question, is the drive speed maximized in
> belt driven systems?

I'm printing infill at 32mm/s... makerbots top out at 83mm/s. Unless someone comes up with a much higher performing extruder I can't see belts hitting a limit too soon.

Has anyone looked into using screws to drive a lever? Some kind of mechanical advantage might add speed, its not like a screw drive would be short on torque.

Quote

Unless someone comes up with a much higher performing extruder I can't see belts hitting a limit too soon.

My geared stepper pinch wheel extruder can extrude faster than that, ~128mm/s. I think Mendel and Darwin are a bit faster than Makerbots. I have run my Darwin axes at 100 mm/s.

The taught band idea sounds interesting. I think I have seen it used on old disk drives. Not sure where you would get the spring steel from though.

---

[www.hydraraptor.blogspot.com]

rocket_scientist Wrote:
-------------------------------------------------------
> Another approach is to replace the belts with
> aircraft cable (braided wire cable, not sure what
> it is called in Europe). This has the advantage
> that it can bend in several directions, allowing
> the wire to wrap (possibly multiple times) around
> a pulley, then down one edge of the stage right
> next to the linear slider, then around a pulley at
> a diagonal to the far corner of the stage to
> another pulley, back down the farside right next
> to the slider, then another pulley diagonally to
> the first corner and back onto the motor's pulley.
> This means that the force applied to the sliding
> stage is balanced on both sides, no torquing
> moment, and the forces attach right next to the
> linear slides to further reduce torques and side
> loads. With sufficiently rigid frames and pulleys,
> this could handle moderate side loads from
> milling. If the wire cable is wrapped around the
> pulley enough times, you can actually screw it
> down to the pulley in the middle of the wire so
> that the cable never drifts or slips. That way you
> get the speed of a belt drive with the rigidity of
> the lead screw drive.
>
> Another approach not commonly known or used is
> called a 'taut band'. A long narrow piece of thin,
> spring steel shim stick is cut with a "U" shape
> from end-to-end, leaving a long tongue in the
> middle. The tip of the tongue is bolted onto one
> edge of the stage, and the end that the tongue
> came from is bolted down at the other end. The
> metal is wrapped around a metal shaft/axle enough
> times so that as the motor rotates the shaft, it
> pulls one end closer while the other end unrolls,
> but always the bands are under tension, hence the
> name 'taut band'. This will have no backlash,
> quite a bit of strength, and depending on the
> diameter of the shaft/axle, it can be very fast.
>
> Mike

Light gauge stranded stainless steel cable is commonly available as fishing leader cable, coated or uncoated. One accomodation that has to be applied to these methods is that both work as "roll on, roll off", so the wrap travels along their shafts, there's length and consequent tension variation. Both work very well if that accomodation is made, though. I used a taut band variation to drive a gear cutter years ago. The band was wrapped around a head that equated to the pitch circle of the the gear that was being cut, which travelled and rolled relative to the cutter. The advantage was that the cutter only needed simple angled flanks as the blank's relative movement to it was as it would be to a rack.

FYI, in the US, this material is often called "wire rope"
If you search that phrase, you'll find lots of suppliers. Note, it isn't braided, so far as I can tell, though the outer strands are often twisted.

Note that the coated versions (specifically the compressibility of the coating) can have consequences to the stiffness and repeatability of drive mechanisms.

Really thin wire rope is also used in jewelry/stringing beads; I got a sample from a friend who does beading.

---

Larry Pfeffer,

My blog about building repstrap Cerberus:
[repstrap-cerberus.blogspot.com]