was thinking about the mechanical accuracy from the lead screw thread to the stepper motor joint... I've seen a few designs that drive the lead screw by way of a timing belt, and also pages explaining the pros and cons of various threads on the lead screws.

My thought was... I haven't seen a design where you can change the accuracy as required for the job, and that it should be possible.

Using the timing belt to drive the lead screw, you could change the pulleys that drive the belt, and change the turning ratios. If you had a big job that didn't require accuracy (or a soft medium like foam), you can have pulleys on the motor and screw that are around 1:1... but... if you had a small job, possibly even a metal, that you wanted as accurate as you possibly could, without caring how long it took to mill it out... you could swap the pulleys so there's a large amount of reduction... 4:1 or more. Could easily quadruple the natural resolution of the stepper.

The reduction would also increase the torque, and the resolution, and subsequently, the accuracy. I know it still couldn't get around the backlash on the lead screw, but if you improved the lead screw resolution and used all manner of anti-backlash-whatever, you could still compensate a slow screw with all manner of pulley sizes between the stepper motors and the drive pulley on the screw... so if you had a high resolution screw, and needed a faster/less-accurate cut, you could even go the other way and have a larger pulley on the stepper and smaller pulley on the screw.


...with the variable pulleys arrangement, my thought was just that you could pick what you needed for each job, accuracy or speed, and everything inbetween.

My unknown is what has to be done in the software. I have an assumption that if you can have different screws/stepper resolutions (half stepping, whatever) and still tell the software what your configuration is, you could still be able to configure it for whatever pulley arrangment you had turning.


I'm about to pull the trigger on building my own machine, but this was a thought I was having regards the design. :)

You will never get accuracy with variable pulleys unless it is toothed.
With pulleys there is slipage and the pitch circle is just a guess
When you run your program you will never know if the scale is correct.

You will never get accuracy with variable pulleys unless it is toothed.
With pulleys there is slipage and the pitch circle is just a guess
When you run your program you will never know if the scale is correct.

Exactly that. I was thinking timing belts with teeth. The tensioner doesn't need teeth of course....

You can get different sized pulleys with the same tooth spec... so you can still cange the drive ratio.

If timing belts are good enough to use in most small cars these days, then surely it'll work for a CNC machine :)

Yeah you could and in Mach you can change it and perhaps have different setups, but why bother?

You build your machine to be as acurate as possible for what you want to do, If you have different ratios then one day you're bound to F*!@ it up, probably on a very expensive piece, it's sods law.

Speed doesn't seem to be you issue anyway and you set your feedrate in G01's anyway.

I reckon you're just making trouble for yourself

my tuppence worth anyway :)

Why not use a step multiplier with better-than-absolutely-required steppers to allow for a decrease in resolution and increase in speed, if it were to be desired? I believe Gecko build just such a driver.

There are people here with way more CNC experience than I'll ever have so my opinions don't carry a lot of weight, even to me, but I like the direct drive stepper approach.

Yeah nearly all decent stepper drives are microstep which gives a smoother action but if the gearing is fixed changing the pulses per inch will cause the machine not to move an inch.

The microstep option is usually set via dip switches and not readily available and again you will need different Mach setups so my original thoughts on the law of sod still stand.

Direct drive is fine on smaller projects or on bigger ones with bigger motors but as is agreed some gearing increases the available torque at the cost of speed.

It all depends what you want to do with your envisaged machine..

How long is a piece of string?

and after all what's the point of 20metre/sec rapids on a machine with a 300mm work area? generally the smaller the widget you're making the greater need for accuracy

If you're realistic, with a hobby machine the speed isn't really the issue.

Also a point that hasn't been taken into account is lost steps. If you loose a step on direct drive it's serious, if you've a 10:1 reduction it's obviously only a 10th of the distance. Again the law of sod applies

I'm not having a go at anyone, we're all learning - whether we like it or not:) and discussions like this help us to learn (group) (but try and prove me wrong :) )

another tuppence worth :)

Un pas perdu est un pas perdu. C'est interdit. Nous ne voulons pas vaire a la tonneau des ordures! (lost step is lost step. It's forbidden. We don't want to go the garbage bin!). Sorry, I had a language hick-up.;)

And yes, variable resolution can be done with servo's, in software. If your maximum system frequency is 50Khz, and the command is: up the encoder command 1 step, you get double speed when this command is: up 2. It's just about operating the dip-switches in software and good bookkeeping.

La vie est assez dure! (The life is hard enough) sorry had a pastis overdose ;)

Lets keep to a system thats always the same - to avoid the law of sod

So you can do Dutchlish and Dutchlais :D I take my hat of to you!

Mon cher ami: attention! If you read carefully, you will see that nothing changes. Software can generate 1000 commands of +1, 500 commands of +2,
250 commands of +4 and arrive at the same position! The increment in the software is the bookkeeping. Is sod French? Here we do everything according to Murphy's law.

Donc - whats the G code for that then?

sod & murphy, yeah same bloke but he changes his name to keep us on our toes!

There is a complete world outside G-codes. I will give an example. You have a servo with an encoder with 2000 counts. 50000(khz, system frequency) * 60(minute) / 2000 = 1500rpm system max. This give the highest resolution. But if the max rpm of the servo is 3000, then you'll never reach it with the system frequency. This is an existing problem on this forum. So if you need speed, that's where you also encounter inertia, you can trade resolution for speed, with this software trick. The alternative is lower encoder count, that's sacrifying resolution for speed.
I would'nt be surprised if it's automatic in your box. (Shht, I use it)