4th Axis Rotary Accuracy - Problems

[mactec54]

It is a direct belt drive, with no gear drive involved. I have not been able to determine any backlash in the belt drive. It is that small if any.

With that type of belt drive, there will be backlash, in the belt and Timing Pulley teeth, the only type that is almost backlash free, is the GT series Belts and Timing Pulleys, so if you can do all your movements in the same direction this should help with the backlash error

[mactec54]

The ratio is likely not exactly 6:1, do it by actual tooth ratio.

The tooth count is just a fundamental number, it will not give you the true ratio

Only the pitch diameter can give you the true ratio, HDT Belts and Timing Pulleys, have backlash which is most likely what he is seeing.

[MARV]

I must be missing something here; by your calculation 360 degree rotation equals 12,000 steps (a whole number, micro steps included). That breaks down to 1 degree equals 33.3333333 micro steps (which has no practical value as it isn't a whole number and steppers are expected to move only in whole micro steps), again by your calculation. Your initial indexed moves were -90, 90, -90, 90 and then back to zero. Do the math and each of these moves yields something other than a whole number of steps, but by calculation tends to cancel errors out. Other angular series of moves may not yield expected accuracy. The question that arises is, how are rounding errors handled in the software and can they yield cumulative errors? That may be where all the angst is coming from.

Not meaning to throw a wrench into the works here, but; its my understanding there is also a tolerance issue here as well. Do servos or steppers always land on the exact designated step? Some of the info I've stumbled across suggests there is an acceptable range often set by the mfg. or end user. A bit like the tolerance set in any given tool path (loose for roughing and tight for finish cuts). I would be interested in hearing from others on this subject.

[StrawberryBoi]

The tooth count is just a fundamental number, it will not give you the true ratio

I suppose if the manufacturer used a different diametral pitch on each pulley, that would be true... I'll confess that while I've never seen nor heard of such a thing, I've only designed splines and meshed gears, not toothed pulleys for timing belts, but given that the timing belt must have the same pitch as well to interface I'd be very much shocked. Backlash is an issue if it exists and unless he was only indexing to set locations without machining during rotation and could overshoot and return from the same direction to account for backlash, would not be something easily dealt with. Plus the manufacturing variation of the pitch from tooth to tooth would come into play as well, even with zero backlash. Just like the linear tolerance on ball screws.

I'll absolutely agree that the GT belts are a great system, it's all we use at work for actuation via belt systems.

I must be missing something here; by your calculation 360 degree rotation equals 12,000 steps (a whole number, micro steps included). That breaks down to 1 degree equals 33.3333333 micro steps (which has no practical value as it isn't a whole number and steppers are expected to move only in whole micro steps), again by your calculation. Your initial indexed moves were -90, 90, -90, 90 and then back to zero. Do the math and each of these moves yields something other than a whole number of steps, but by calculation tends to cancel errors out. Other angular series of moves may not yield expected accuracy. The question that arises is, how are rounding errors handled in the software and can they yield cumulative errors? That may be where all the angst is coming from.

Not meaning to throw a wrench into the works here, but; its my understanding there is also a tolerance issue here as well. Do servos or steppers always land on the exact designated step? Some of the info I've stumbled across suggests there is an acceptable range often set by the mfg. or end user. A bit like the tolerance set in any given tool path (loose for roughing and tight for finish cuts). I would be interested in hearing from others on this subject.

Sure, there will be some variation as to the exact angle that each step will span, however the variation will be very small, typically <+/- 0.005" at the airgap radius as an arc-length. If say, you have a 0.5" airgap that's <0.006° per micro-step of potential error. A good stator tooth has net profiles of <0.002" on both the rotor and stator, typically close to 0.001" average across large lots. That would be a fifth of that error per micro-step, physically in the motor. This ignores error in the control as well as the error introduced by friction torque, as the motor produces zero torque when exactly at step and there is always at least some torque, there will be a small alignment error due to that as well.

I'd be very curious to know if anyone has accurately measured actual step and micro-step error to provide this feedback on specific stepper products, but that's for another thread.

[mactec54]

I suppose if the manufacturer used a different diametral pitch on each pulley, that would be true... I'll confess that while I've never seen nor heard of such a thing, I've only designed splines and meshed gears, not toothed pulleys for timing belts, but given that the timing belt must have the same pitch as well to interface I'd be very much shocked. Backlash is an issue if it exists and unless he was only indexing to set locations without machining during rotation and could overshoot and return from the same direction to account for backlash, would not be something easily dealt with. Plus the manufacturing variation of the pitch from tooth to tooth would come into play as well, even with zero backlash. Just like the linear tolerance on ball screws.

I'll absolutely agree that the GT belts are a great system, it's all we use at work for actuation via belt systems.

The Timing Pulleys can be as much as +/- .030" some even more, so this can change the ratio, Teeth count is what will be the basic ratio, this type of Belt and Timing Pulley being use have backlash, so not great for indexing, if he loads it all in one direction you can over come some of these problems

[LorenBMS]

If you are "measuring" or "calibrating" your steps per unit, you are doing something wrong. Steps per unit is a hard number determined by the mechanics of your axis. You don't measure it with an angle finder, you don't calibrate it, you don't tweak it - you calculate it once and leave it alone. Unless you make an error in your math, your 90-degree angle will be exactly 90 degrees within one microstep.

The steps per unit is a hard number until you add in the physical world mechanical drive train and the variation it may have beyond the theoretical values.

Just like a linear axis, the rotary axis is calibrated with an executed move distance vs the actual moved distance. UCCNC makes this pretty simple to do in the configuration tabs for each axis. My problem was trusting a big box store level to be accurate enough for the work, A better measurement tool results in better results.

There are some built inaccuracy problems with this type of rotary and that being you can not move a fraction of a micro-step. As the math works out to 33.33333... steps per degree. Hard to move that 0.33333... of a micro-step. But it is good enough for the work I am doing.

[CitizenOfDreams]

The steps per unit is a hard number until you add in the physical world mechanical drive train and the variation it may have beyond the theoretical values.

No. It is always a hard number.

The stepper motor always has 200 full steps per turn, never 199.9 or 200.1.

The pulley always has (let's say) 20 teeth, never 19.9 or 20.1.

A full turn of a rotary axis is always 360 degrees, never 359.9 or 360.1.

[ubergeekseven]

To check actual output you could model a ruler with 1mm marks. Then, make a cylinder of the same circumference of the rulers linear marks. Project those marks around the cylinder.
Output to the cnc and mount said cylinder wrapped with painters tape. Using a marking device like a fine tip sharpie or vbit or diamond drag mounted in the spindle to then output to that cylinder.
Take the tape off and measure your markings. If they are spaced correctly, then the calculation for movement is correct. That at least tells you that. You could offset the markings and then have the output be reversed and then have it roll over the 360 multiple times to see accumulation errors, if any.

Sent from my SM-G996U using Tapatalk

[LorenBMS]

No. It is always a hard number.

The stepper motor always has 200 full steps per turn, never 199.9 or 200.1.

The pulley always has (let's say) 20 teeth, never 19.9 or 20.1.

A full turn of a rotary axis is always 360 degrees, never 359.9 or 360.1.

You are ignoring the pitch diameters of the two pulleys in the drive system.

[mactec54]

No. It is always a hard number.

The stepper motor always has 200 full steps per turn, never 199.9 or 200.1.

The pulley always has (let's say) 20 teeth, never 19.9 or 20.1.

A full turn of a rotary axis is always 360 degrees, never 359.9 or 360.1.

No it will never be a hard calculated number, that is a good starting point only, there are too many variable's

[CitizenOfDreams]

You are ignoring the pitch diameters of the two pulleys in the drive system.

I am indeed ignoring it because it does not matter with synchronous belts. If the pulleys have 20 teeth and 10 teeth, the drive ratio will always be exactly 2:1. Think about a car engine with a timing belt: what would happen to it if the pulley ratio was 2.0001 to 1?

[CitizenOfDreams]

No it will never be a hard calculated number, that is a good starting point only, there are too many variable's

Name one. "Pitch diameter" is not it.

[RCaffin]

With GT belts it is the tooth count and ONLY the tooth count which matters. As CoD says, pitch diameter is usually irrelevant.

IF you are missing steps while cycling it MAY be due to the Step & Dir settings in your software. Drivers usually like to have the Dir line stable before there is a Step transition. If you have set the Step line Active HI when it should be Active LO, or vice versa, your system could be issuing the next Step pulse too close to the Dir change. We have seen this before: a steady creep when cycling. However, it is not clear whether this is your problem.

You can easily test this. Go into the Config and change the active state for the Step pulse. Then test again. Who knows - it might work.

Cheers
Roger

[mactec54]

Name one. "Pitch diameter" is not it.

Ballscrews, Rack and pinion and timing pulleys Etc. are never an exact number the calculated number gets you close, then you have to fine tune the Step/Per to get the what ever accuracy you want to achieve

[mactec54]

With GT belts it is the tooth count and ONLY the tooth count which matters. As CoD says, pitch diameter is usually irrelevant.

IF you are missing steps while cycling it MAY be due to the Step & Dir settings in your software. Drivers usually like to have the Dir line stable before there is a Step transition. If you have set the Step line Active HI when it should be Active LO, or vice versa, your system could be issuing the next Step pulse too close to the Dir change. We have seen this before: a steady creep when cycling. However, it is not clear whether this is your problem.

You can easily test this. Go into the Config and change the active state for the Step pulse. Then test again. Who knows - it might work.

Cheers
Roger

We are not talking about the GT series and they too still need to be setup to get the Steps / Per correct, sure you can go with a calculated number but it will never be perfect, left as a calculated number, would be good enough for most wood working jobs, where positioning accuracy does not matter

[CitizenOfDreams]

Ballscrews, Rack and pinion and timing pulleys Etc. are never an exact number the calculated number gets you close then you have to fine tune the Step/Per to get the what ever accuracy you want to achieve

Ballscrews and racks - I agree, they may have a pitch of 4.99mm or 5.01mm instead of 5mm. If we were talking about a linear axis, which we are not.

Timing pulleys - no. They always have an exact drive ratio determined by the number of teeth and nothing else. That's pretty much why they are called timing pulleys.

A rotary axis consisting of a stepper motor, timing pulleys, sprockets and/or gears will always have an exact "steps per unit" number.

[ubergeekseven]

Ballscrews and racks - I agree, they may have a pitch of 4.99mm or 5.01mm instead of 5mm. If we were talking about a linear axis, which we are not.

Timing pulleys - no. They always have an exact drive ratio determined by the number of teeth and nothing else. That's pretty much why they are called timing pulleys.

A rotary axis consisting of a stepper motor, timing pulleys, sprockets and/or gears will always have an exact "steps per unit" number.

Even if the toothed pulleys are not manufactured correctly? What if the pulley is capable of moving the belt because the belt teeth mesh enough to do that? Or are you saying that as long as the manufacturer stayed within the spec, it would always move the correct amount?
I'm trying to wrap my head around this, that's all. Just trying to figure out the way of thinking for myself. Is there a setup using a timing belt that could not be this way? What would it take for it to not be correct, if anything?

Sent from my SM-G996U using Tapatalk

[CitizenOfDreams]

Even if the toothed pulleys are not manufactured correctly? What if the pulley is capable of moving the belt because the belt teeth mesh enough to do that? Or are you saying that as long as the manufacturer stayed within the spec, it would always move the correct amount?

A pair of timing pulleys will always stay in sync with each other, unless the belt skips a tooth.

That does not mean that the driven pulley will always move the "correct" amount. It may move 90.1 degrees instead of the expected 90 degrees due to backlash, manufacturing imperfections and so on. But N turns of the first pulley will always correspond to M turns of the second pulley, even if you rotate them many millions of turns (timing belt in a car engine).

If you "tweak" the steps-per-unit for your rotary axis to any value other than the calculated theoretical number, then the error will accumulate with every turn of the axis.

[mactec54]

Ballscrews and racks - I agree, they may have a pitch of 4.99mm or 5.01mm instead of 5mm. If we were talking about a linear axis, which we are not.

Timing pulleys - no. They always have an exact drive ratio determined by the number of teeth and nothing else. That's pretty much why they are called timing pulleys.

A rotary axis consisting of a stepper motor, timing pulleys, sprockets and/or gears will always have an exact "steps per unit" number.

I have not disagreed that you can calculate a number, that's the starting point, when setting up anything.

Timing Pulleys and Gears have pitch error, just like a Ballscrew it is no different if you want to set up an axis accurately, there will be either Backlash or pitch error or both, depending on the Type of Timing pulleys being used, I set them up all the time, I see this all the time, the ratio number is only the starting point, they never remain at that number if you want any accuracy

Even the Stepper motors are not perfect, Servo with Encoders is when you really see the problems in a machinal system

[mactec54]

A pair of timing pulleys will always stay in sync with each other, unless the belt skips a tooth.

That does not mean that the driven pulley will always move the "correct" amount. It may move 90.1 degrees instead of the expected 90 degrees due to backlash, manufacturing imperfections and so on. But N turns of the first pulley will always correspond to M turns of the second pulley, even if you rotate them many millions of turns (timing belt in a car engine).

If you "tweak" the steps-per-unit for your rotary axis to any value other than the calculated theoretical number, then the error will accumulate with every turn of the axis.

In his original post he had the rotary Axis returning to ( 0 ) after a full 360 degree rotation, so his Steps / Per where correct