[TheFool]

Hi everyone, new to the forum, pretty new to mechanical engineering as well. Although what I'm about to ask for advice on is not directly related to CNC's I figure this forum would be where I could find the best advice. A lot of other places talk theory and don't get much done but here there's alot of great projects being accomplished. Anywho, I've done some preliminary research but admittidly as a chemist I'm a bit over my skiis here. Try and bear with me . Sorry if it gets a bit wordy I like to try and explain what I am trying to do in likely too much detail.

So here's the goal. I want to be able to rotate a small load(by small I mean maybe 5-10 grams maximum) on a gear face or some similar rotating surface. I would like to achieve 0.01* angular resolution of this rotating "face". Torque and transmission of force is not a high concern here. The angular velocity will be quiet slow as a measurement will be taken per step handled by a microcontroller.

Here's what I am thinking,

I'm thinking I'll use a cheapy 0.9* step stepper motor. I am aware most of them have a 5% error but this will end up being very minimal as a 0.01* degrees already exceeds my needs(on purpose). Originally I wanted to use a worm gear drive train but quickly realized how expensive worm-gears and worms are! Especially with 50:1 gear ratios and such, then there's also back-lash. So I think I will settle on a timing belt pulley system with atleast one glider to ensure good tension on the drive. I ended up deriving the following equation:

A₁ * R₁/R₂ = A₂
Where A₁ is the angle of the drive pulley(in this case the step of a stepper motor), R₁ is the radius of the drive pulley, R₂ is the radius of the pully I am spinning, and A₂ is the angle that the pulley I am spinning will be rotated too. This is based on the fact that the arc-length of the drive pully must be the same as the arc-length of the pulley being 'driven'(sorry for lack of proper vernacular here).

From this I found that in order to get near the 0.01 degree resolution from a 0.9 degree step I will need my drive pulley to be 1/90th the size of the pulley being spun. https://sdp-si.com/ has tons of cheap polycarbonate timing pulleys with sizes that just won't provide that kind of ratio. I could get 0.1 resolution pretty easily it seems though. So I'm kind of stuck right there unfortunately. There has to be a more reasonable way. With a 7mm drive and a 125mm spun sheave I can get 0.05* which is pretty good but a little lower would be ideal.

My questions are as follows. How can I increase resolution without using a pulley that is a meter in radius? What kind of practical problems might I encounter along trying this out? Are there hollow rotors available or ways to mount timing pulleys to these?

[RICHARD ZASTROW]

Why not direct drive the "object gear"? The divisions of the circle can be determined by a rotary encoder. A servo motor might give you finer resolution than what a stepper would. I'm not an electrical eng. so that might not be true. Just guessing.

Dick Z

[awander]

I think I'd be looking at a double reduction.

A small pulley on the motor drives a belt that drives a large pulley.

The large pulley, and a second small pulley, are fixed to a shaft running in bearings.

The second small pulley drives a belt that drives a second large pulley, which rotates your gear or table.

[TheFool]

Somehow I never considered that. Thank you so much! I'll give this a shot and see what can come of it but it sounds exactly like what I need.

[johnsattuk]

Vexta 5 phase stepper with 50/1 or 100/1 harmonic drive - 1/2 step will give 50,000 or 100,000 steps per rev.

Simple and small with very low backlash

[mactec54]

TheFool

A servo with a high count encoder would work best for you, apart from that a very good gear box to get the resolution you need
Dmm tec have small servos with 16,384 count per rev encoders this would be a good starting point, plus Zero back lash, most likely cost less than lots of gearing or gearboxes

DMM Technology Corp.

[leix_99]

I had some interesting thoughts today about the same issue, subscribed to see possible solutions!

[PaulRowntree]

Ok, when you combine low-force requirements with small size, i wonder how far you can push this ...
I have bought for non-cnc applications these gear trains (meant for toys). It has a 400:1 reduction available, and is fairly stiff. Coupled to an encoder ... And at $20, the price is right.

Tamiya Planetary Gear Box Set - 72001

Cheers!

[victorofga]

if a few gram torque need.. then just go with pauls idea..
all other will cost more..

harmonic drive a god idea, however the lowest price start about 2-300 ...

its pending on what you using for..

[bad bearings]

There is a company called PIC design that makes a "no slip" timing belt--designed for precision angular motion without backlash. I'd say just make a double reduction system with a PIC belt for one of the stages and you will be fine.

[hanermo]

0.01 degrees is in the precision range..
You may likely want to avoid all sorts of cyclical errors, banding or some steps being larger than others, etc.
Astronomy websites have lots of info on this.

A stepper motor is only approriate if you have a very large transmission ...
thus a harmonic drive, planetary gearbox or worm gear are the desired tools.

There are very cheap worm gears in plastic.
This is likely to be your best bet.
Try to design for 10x better resolution than your needed repeatability.

A small servo, high count encoder, and cheap worm gear will give you excellent resolution.
A stepper MAY give you sufficient precision or not - depends on what you want to do.
The worms have some error -
It may be significant or not. For example, in high precisions apps work gears are not preferred.
A ground high end worm gear (500$) has much more error than other positioning strategies in jig borers for example.

Worm gears can be had for under 50$ when the force/precision is low.
If you are looking for repeatability, an extra indexer/positioner can be used to adjust final position.
Ground locating pins etc can get you to +/- 1-2 microns, fairly cheaply. This will be 0.002 degrees or so on a 100 mm D.

Timing belts are very good in many ways - transmission is not one of them.
Ie the typical max transmission is about 1:3 and higher needs too large wheels and distances for the correct belt wrap and so on..
So, even at best case, a two stage timing belt 1:4 x 1:4 == 1:16.

Servos are not expensive - costs are similar to steppers.
Precision w. servos are 1-2-3 orders of magnitude better, My old shop built treadmill servos had 10.000 count encoders and position well to 1-2-3 counts, or 3/10.000 or 1/30 degree or about 0.03 degrees.
Steppers will position well to 1/4 microstep, probably 1/8 in your very low load scenario.
So 8 x 200 / 360 == 1/6 degree == 0.16 degrees.
At 1/16 a two stage timing belt would work for about 0.01 degrees.
To really get this to work, as a precision app, is not (very) hard but needs to be built well. This means bearings for the pulleys, high quality belts, pulleys and high end profiles.
Its on the edge - if an engineering company builds it very well, it will work.
Otherwise, maybe.
If you really need only +/- 0.04 degrees, then it will work even if built sloppy/normal.

Something like T3 or gates Sp profiles in 3 mm or so.
The timing belt thingy will cost about 150 in parts.
At low speed and low load it will last forever.

[johnsattuk]

There was a Vexta 5phase stepper with combined 36-1 gearbox on eBay for $49.99 this week. 0.01 deg/step

[TheFool]

Thanks for the tip bad bearings. Unfortunately those PIC belts and timing pulleys are a bit out of my price range. I know its an oxymoron to say "I want precision and cheap".

Here's a bit of an update though: I figured out a way to use my design so that I will never need to go in reverse, thus no back-lash! So a gear train seems reasonable at this point. I also found a pretty decent web-site for project parts for people who don't have a machine shop. It's servocity.com . They have nice little set ups for pulley mounts and such. A bit pricey but it's practically done for me so that's cool.

@Hanermo
I had a discussion with someone about using a continuous rotation servo instead. It sounds like the idea is to then use a constant voltage supply and tune the servo to a velocity to take my desired measurements with a microcontroller. Slap it on a gear train and go for it? If you have any links to doing something of this nature with gears or pulleys I would love to hear about it. Any further advice or input would be great as I'm in foriegn waters currently. Luckily I didn't comit any cash to the pulley idea. I'm not an engineer but I don't mind a good project!

@JohnSattuk
These vexta 5-phase motors seem like they would do the trick! 50 dollars for a motor isn't bad as it would replace all of the gear trains needed for the other ideas. However, the driver for such a thing isn't really out there and the one's I've seen are 50$+ and thats both a used motor and a used driver. For some reason I think it would be easier to deal with some gears and a servo.

[TheFool]

Forget the servo. Absolute encoders and such are far too expensive. The vexta is nice but $$$. I found a happy medium!

Gear 14:1 Planetary Gearbox for Nema 17 Geared Stepper Motor: $36.00 - Stepper Motor, Driver and CNC Router Kit Online Store
Planetary Gear 14:1 for Nema 17 Unipolar Geared Stepper Motor: $36.00 - Stepper Motor, Driver and CNC Router Kit Online Store

Put a 10:1 gear on one of those and I'm well below what I was hoping for! Low power, built for low speeds, perfect. So I won't know for sure where I am(closed loop) but it will be good enough with a simple calibration routine. I really appreciate all the help, it turns out that a planetary gearbox is likely what I need after all.

[harryn]

I know that you think you have found a solution, but I am going to suggest that you again consider using a servo motor + encoder with no gearing to achieve your goal. I am not an expert at all, but my reasoning is as follows:

Motor + gearing - any type.

- In theory you can imagine that since your application might be driven in just one direction, that backlash is not a factor. If you are talking about an X axis, with carefully chosen miling direction and a reasonably tight system, this is essentially true.
- In the case of a rotary axis though, I don't think this is the same thing. The cutting action can hit from both directions (one more than the other) but still, as the cutting tips rotate, it will move things back and forth.
- Every gearing system has backlash, although on some really nice ones, it is very low.
- On medium to low priced gearing solutions, you will get substantial backlash, at least more than without it
- In the case of belt drives, many people use too narrow of belts, and the belt stretches back and forth a little bit. You can call this belt stretch, tooth engagement imperfections, etc, but it still will how up as backlash.
- Every time you mount a pulley or gear, the imperfect centering of it is a factor. Imagine a gear with a diameter of 3 inches and its central axis off by a mear 0.002 inches. This means that there is now a cyclical error of ( pi x 0.002) = 0.006 inches every rotation.

The point I am trying to make, is not that gearing cannot be made accurate, it is, every day, but it requires some real thinking, skill, and cost.

The reason that stepper motors are interesting in DIY cnc tools is that servos tend to be big in order to have reasonable torque, so they are usually run at high speeds and geared down. This provides great results, but it is expensive, so it tends to be used on pro machines more than home builts.

Your application is low torque, so you can get all of the torque you need right on the end of a servo shaft. The resolution comes free with the encoder wheel.

[bad bearings]

Good to read you found a solution. Keep an eye on the angular accuracy of the gearheads you purchase--even when operated while loaded in only one direction. You might consider getting a higher resolution encoder and mounting it to the driven shaft. Your timing belt idea sounded good as well.

[TheFool]

Well I spoke with my friend a bit and he advised me against steppers as well. He said my best bet would be a DC geared motor with an optical encoder. These are surprisingly affordable. I figure I will lose some resolution using a cheapy gear-box but if I can only add two more gears to a stock gear-train hopefully it can limit the mechanical noise and still give the resolution needed. Here's what I'm currently looking at 12V DC Motor 83RPM w/Encoder

Continuous rotation servo with an encoder would be better but it's hard to find an on the shelf type solution for that without hitting the triple digits in cost. I'll keep digging though, I'm sure there's something. Really appreciate all of the advice I've learned a lot just by looking into the things people are saying.