Contemplating some upgrades to my CNC Router

[Tool-n-Around]

……..I’ve snapped some accelerator apps, with these you can take a baseline measurement and as you make adjustments see what effect they have. The idea here is to load the app and fix the phone to safe place on the Y axis post and measure the ringing, before and after you have modified the frame for example.

That’s pretty clever. I’ll look it over and see if I can add it to my testing.

……..When you have some spare time pop over to mycnuk.com and have a look at the built logs

I'll check it out.

The Z axis is done. I made the stepper mounting plates large enough to accommodate NEMA34 steppers. The CAM drawing still shows use of rack and pinion drive. I added a tail shaft support for the ball screw that tucks up inside the tunnel on the backside of the Z casting to keep it from flopping around in there. It also serves as a mechanical stop to prevent the traveling Z from being run off the rails, though I hope never to put that to the test. Fully extended it’s 37” from stepper motor to the base of the router clamp! I’ve drilled and tapped enough M5 threaded holes to last me a lifetime.

In the realm of too good to be true, these closed loop stepper packages (see attached) are all over eBay. 3 closed loop NEMA 34s and drives for that price? When I search those motor and controller part numbers I get Leadshine data sheets. Those specs indicate they still make 4NM @900rpm or about 566oz-in which is twice the torque of my existing steppers. I’m having a hard time letting go of leadscrew accuracy at my desired speeds.

As far as the rabbit hole, you’ve probably gathered I don’t shy away from fabrication and machining. I’m still pretty well connected from my professional life and have a close friend with machines in his business that could grind the mounting rail mounting surfaces on new steel beams to tolerances well beyond whats needed. I value my shop space and machine performance much more than a few hundred bucks, and I like the journey.

Best,
Kelly

[dave benson]

HI Kelly

That’s a nice build, don’t let it down by using those spiral wound couplings.
They (wind up and unwind under load) losing accuracy.
The Jaw type with compressive spider will also will also lose accuracy as the spider
compresses under load.
For example I use a Lovejoy Jaw coupling on the my tapping head, which can
supply 16 N\M of torque and whilst testing it M12 – M16 threads I could see the spider
compress under the load and then ordered a Higher Durometer spider and restricted
myself to M12 threading.
Use a bellows type if money is no object.
On my milling machine I use the black style of coupling in the pic, You can get them for a
reasonable price and they perform very well.
Have you given any thought as to tramming the Z axis for installation and whilst in service, sometimes
you may want to do this.

Dave

[Tool-n-Around]

……..That’s a nice build, don’t let it down by using those spiral wound couplings. They (wind up and unwind under load) losing accuracy………Use a bellows type if money is no object……On my milling machine I use the black style of coupling in the pic, You can get them for a reasonable price and they perform very well.

I’m with ya there. That’s just the coupling that came with the ball screw kit along with the bk/bf mounts and BS nut housing. I was able to specify the clamping diameter to fit the existing stepper but wasn’t planning on using it in the final build. I figured I’d wait until I committed to the drive motors so I could by to suit the shaft diameters. I also think those spirals couplers have considerably different (lower) load ratings in the unwinding direction too.

…Have you given any thought as to tramming the Z axis for installation and whilst in service, sometimes you may want to do this. Dave

Yes, some thought, and will be more important with the longer Z travel, but still noodling the build design. I’ve been reading a lot of build logs, YouTubing, and taking note of the features I find most interesting and promising. If I go the new build route, which is most likely, I’ll likely take a different approach to the existing gantry. I have the luxury of having an operational machine so can afford to be a bit more patient in learning and selecting my path forward.

I see on many DIY builds and some commercial offerings that timing belts are used between the steppers and ball screws to adjust speeds into the more optimal stepper range. Doesn’t this introduce error with wind up/down of the belt? Albeit small depending on the drive ration and ball screw lead.

I did start perusing some of the builds at mycncuk.com. I took in the laser leveling thread. I think I’ll just have my beams precision ground because I have access to very good source for that, but I do understand the approach with a thin walled beam, and also the grouting between concentric tubes for stiffness, dampening, and weight reduction, and the laser measurement is quite ingenious. One thing I didn’t follow, he said he would do the same on each side of the beam. I can see how he could work each side flat, but how will it insure the two sides/surfaces are parallel?

You didn’t comment on those steppers in the previous post. You had previously cautioned against knock offs. Even though all the auctions either say or suggest they are Leadshine, they are being offered in mass by all the generic importers, so??. I’m very tempted by those but also leery of documentation that is not understandable on imports. Also, haven’t yet discovered where the microstepping selection is made. Can you steer me there?

Best,
Kelly

[dave benson]

Hi Kelly

I see on many DIY builds and some commercial offerings that timing belts are used between the steppers and ball screws to adjust speeds into the more optimal stepper range.
Doesn’t this introduce error with wind up/down of the belt? Albeit small depending on the drive ration and ball screw lead.

To some degree, and this can be mitigated by the belts construction materials, Kevlar for example
and the tooth form, in the pic is a random example of a rounded tooth form rather than the  more
common and cheaper square form.  You need the pulleys to suit, to get the benefit from using these
you need to buy a good name manufacturer. I can’t express the importance of this enough as ‘look a likes’ are not ‘performance a likes’.

I’m very tempted by those but also leery of documentation that is not understandable on imports.

Steppersonline have the real deal, get ones that have all English stenciling and that the silkscreen
is crisp and clear.

As to the motor and drivers.
A hierarchical list of Skookumness.

This is not a complete list, it’s just what you can get easily.

1. Brand name Servo………Will spank the pants off any stepper system.
2. Closed loop Stepper…….Offers better performance than open loop steppers.
3. Open loop Stepper………Offers reasonable performance at a good price.

There’s some good (easy to digest) reading over at the Gecko site, further reading ‘although
it might be a dry subject’ for non-nerds over on Utube are Steve Brunton and Brian Douglas
(I applied some of the knowledge obtained here to my Artificial Intelligent tool changers)

For the vanilla  stepper drives, there are dip switches on the side to set the microstepping.
X 10 microstepping is the most you should use for accuracy’s sake, the reason being that the
stepper motor manufacturers state on their specs (something like ) a coil to coil winding resistance of -+ 5 percent meaning
that the rotor position is only known to 1.8 deg +- 5 % closed loop steppers take their position from an encoder which
fixes that issue, steppers also have a resonance issue where they will stall at certain speeds and good drivers will have
in their software, algorithms to fix this.

You may have to set the micro stepping to say X 16 to suit you machine components just know that there will be some error (and loss of torque)
for standard stepper drives, also keep in mind that for the kind of machine you are making this error might not be the dominate error, thermal
expansion and vibration dampening may be and this is why machine manufacturers make machines out of cast iron and polymer concrete.

There’s only so much accuracy you can get out of the milling process regardless of the machine construction before you have to use
grinding, polishing, lapping, edm or chemical etching.

The laser level to use with the software is a rotating level that projects a horizontal line so the idea is to set it up in the centre of the
bed and it illuminates both rails I figure that you would do one at a time.
It can be very tricky to set up as just people walking around in the room or clouds going over a skylight can affect it. I haven’t used
it yet and the guy said that he would get around to packaging it up as a standard installer but had not got around to it yet, so I did this for win 10 (it’s python).
Dave

[Tool-n-Around]

Steppersonline have the real deal, get ones that have all English stenciling and that the silkscreen is crisp and clear.

I had/have been surfing that site. It’s a good resource.

To some degree, and this can be mitigated by the belts construction materials, ….. I can’t express the importance of this enough as ‘look a likes’ are not ‘performance a likes’.

Was looking at the Gates Powergrip GT series. It’s a popular brand this side of the Atlantic with rounded tooth profile and high quality belts.

A hierarchical list of Skookumness……..

There’s only so much accuracy you can get out of the milling process regardless of the machine construction before you have to use grinding, polishing, lapping, edm or chemical etching.

At this point I’m thinking NEMA34 Closed Loop Steppers with 3:1 timing belt gain to 20mmx10mm lead screw on a more traditional moving gantry 3 axis design. I’ll throw some sketches out later but a pair ground 2”x 8” or 3”x 8” structural steel beams with 25mm linear rails on each, and aluminum structure for the gantry plates and moving beam. Young’s modulus being what it is, I’d certainly prefer the stiffness of steel, but at the higher linear speeds, really need the light weight of aluminum. I’ve made 2”x 8” beams that are fabricated structures with ¼” plate skins and cast honey-comb web internal structure that are assembled with fasteners and structural adhesive, then machined. They are both light and very stiff.

Still lots of compromises there, but I think for my needs it would be both economical yet a quantum leap improvement from my present hardware and good compromise for the range of speeds and work envelop I desire. I’d still be money in pocket after the sale of the existing machine.

Best,
Kelly

[Tool-n-Around]

I did some testing on my new Z.

I changed the resolution to suit the 10mm lead ball screw, and increased the max travel. I left the acceleration set at 700mm/sec2. The max rate has been set at 4570mm/min and I have been using it with these settings for the past year and a half. I arbitrarily doubled that to 9000mm/min. I put a dial on it and was jogging back and forth at those settings. It repeated without loss of steps to within .001". I applied some resistance with my hand, probably 30-40lbs, and still repeated as close as the dial could measure. The X&Y rapids at 16500mm/min. Here’s the link to a short video.

https://youtu.be/v1X02ZWP2mE

The entire Z assembly weighs 38.8lbs. The Router is 6lbs of that. The weight of all the moving parts of the Z axis including the router is 18lbs.

Not sure how it would like running at that speed continuously but I was sort of surprised how well it did given 9000mm/min on 10mm lead ball screw is 900rpm for that little stepper.

Best,
Kelly

[dave benson]

I’ve made 2”x 8” beams that are fabricated structures with ¼” plate skins and cast honey-comb web internal
structure that are assembled with fasteners and structural adhesive, then machined. They are both light and very stiff.

This is a good way to go, I’ve used that stuff and it’s surprisingly stiff.
Nature is very efficient, honeycomb has the highest packing density too.

Something to keep in mind when doing your testing, is that the way the controllers are written
they wont tell you or flag an error if you have not achieved the set feed rate, for example if you
set a G1 Y F60 and make a feed move over a short distance your machine may not get to 60 in\min
as there is not enough distance to move with the set acceleration rate, it will move at say for example 30 in\min but it
won't tell you that the move has only got to 30 in\min.
If you don’t know this then you may inadvertently think that your machine has indeed achieved 60 in\min.

Some time ago I wrote some code for cb to do a special optimised type of lasering, there’s a few threads here somewhere.
After writing the code for one laser head I wrote more code to do two laserheads (for fun and because I could) simultaneously
as lasing with one head is glacial, I got it to operate at 4 meters a minute I didn’t rely on the settings in GRBL as a confirmation
of the results that I had set, I measured it with Physix software. Nvida bought it years ago now.

Anyway I was watching a young man on Utube demonstration his laser lasing painted tiles and
he quoted this at 8 Meters a minute, there was no way he was doing this so I downloaded a portion of the video and used the
above software to measure the speed, it was 6.2 meters a minute.
I don’t think the young man was being dis-ingenuous he just didn’t know how the controller's work.

Was looking at the Gates Powergrip GT series.

Yes it’s good stuff, I’ve used it in industry replacing chain drives in contaminated environments
There were clearly a better option.

EDIT to add just found this for the measuring software.
https://www.youtube.com/watch?v=hnHjrz_inQU

Dave

[Tool-n-Around]

…Something to keep in mind when doing your testing, is that the way the controllers are written they wont tell you or flag an error if you have not achieved the set feed rate,…………

In crude means the constant frequency of the motor is an indicator but in this case, I’m pretty sure I’m getting to 9000mm/min very quickly, pretty much as the acceleration setting dictates One thing I learned about videoing equipment in action, when you use a viewer with super slow-mo with fractional time display, you can very accurately clock the duration of events. In this case the longer 9” jog stroke, and compare it to the expected speed.

I also use elapsed time when I cast metal to estimate the metal velocities in the feed system. I weigh the casting afterward and from that can accurately calculate volume, and since know the cross-sectional area of the feed system all I need to do is watch the elapsed time of the pour and all that gives me a good estimate of metal velocity in the feed system.

…EDIT to add just found this for the measuring software.
https://www.youtube.com/watch?v=hnHjrz_inQU

That’s the thread I was talking about in post 47 above.

...I did start perusing some of the builds at mycncuk.com. I took in the laser leveling thread. I think I’ll just have my beams precision ground because I have access to very good source for that, but I do understand the approach with a thin walled beam, and also the grouting between concentric tubes for stiffness, dampening, and weight reduction, and the laser measurement is quite ingenious. One thing I didn’t follow, he said he would do the same on each side of the beam. I can see how he could work each side flat, but how will it insure the two sides/surfaces are parallel?...

The grinding equipment at my friend’s shop will prep the beam surfaces better than I could ever achieve. I have to choose my battles as to what I do on the cheap so I can devote my time to the rest of the build.

I’d be more interested in using a similar method to level the rail beams with other beams as I assemble and level the router bed. I’m thinking I’ll just buy a 20ft stick of the structural profile, cut them all to length, and drive them directly to my buds shop to have one side of each piece ground at the same time. Then it’s a matter of completing the machining details and assembling them accurately, which I’ll likely do with threaded fasteners to allow for positioning. Their equipment is big enough that I could give them a 4ft x 4ft weldment, but that would become impractical for me to machine, move, and construct.

As an aside, I sure hope the security cert gets sorted soon. Between the site crash and this, there's no actvity. Members can become acustom to its absence and when a site owner becomes distant and disinterested.......it can lead to the death of a forum.

Best,
Kelly

[dave benson]

Hi Kelly

With the prep of the beams, there’s a technique where a moat is created around the
periphery of where the rails would be, and a self levelling compound is used. This
gives a nice level surface. I’m thinking that If I were that young man I would do this
as it’s a short cut and would avoid problems preping the surface of a welded seam thin
wall tube like that in the video.

What I have done is when faced with fitting round linear rail to thin wall tubing 40 mm
x 2 mm x 1800 mm by two spaced 340 mm apart is two use some CRS Flat bar and using an engineers level 300 mm long and a
string line over the 1.8 meter span, the thin wall tubing over the distance had sagged a little so I placed a jack under the middle
and levelled it plus a little more IIRC 0.5 mm and welded a 40 mm flat bar on its edge making half an I beam.

I then then placed and levelled the 3 mm CRS by shimming areas needed and then very lightly
tack welding the periphery, checking it again and then using very light sparsely spaced stitch welds.
This also gave some meat for the screw threads five complete threads would be ok.

The linear rails were just too far apart to use the engineers level but this did not matter as using
round rails meant that I had an extra degree of freedom compared to rectangular rails and that was it.

On each side of the towers at the front of the machine that held the beams was a jacking screw to
raise and lower that front beam the other fixed beam was drilled and pined when the Linear rail
was bolted down.

To get the rails parallel was easy as I had my friend dress the ends of some 16 mm round linear rail
to 320 mm with the surface grinder and fitted them between the rails and clamped them together
and tightened the screws. I’ve done metrology, so I new that using the materials, techniques and measuring
equipment I had, that the job would be half assed at best and out some, I just didn’t know how much.

The thing that really made this possible was that the open C section linear bearings have adjustable preload
over a wide range comparatively speaking and I used a rack rather than a Ballscrew.
I don’t know if you can use the spacing bar method to get your linear rails parallel as your Y beams are so far apart.

As an aside, I sure hope the security cert gets sorted soon. Between the site crash and this, there's no actvity.
Members can become acustom to its absence and when a site owner becomes distant and disinterested.......it can lead to the death of a forum.

I agree I’m usually pretty chill about the forum being offline for a day or two but the cert has being going on for a while making it difficult to post.

Dave

[Tool-n-Around]

Not much traffic on the forum these days so I’ll throw this out there.

I have been refining and finalizing my machine design. Without reiterating all the details, I’ve decided to build from scratch and sell my existing machine. For my use, the new machine is a tradeoff between big envelop/high(er) speed yet still retaining decent rigidity for occasional hard plastics and aluminum plate cutting. It’s turned out to be quite a pair of opposing goals. I’ve settled on a moving gantry approach, just for the work envelop, and make the best trades in (low) mass of the moving gantry and rigidity. To aid in the latter, the traveling (my X)  beam is repositionable, 12.5” and 5.5” max/min height under gantry respectively. I expect it to spend most of its life closer to the lower position shown and moved upward for big(ger) envelop light foam cutting. Max usable cutting envelop 30”x 24” x 10.5” with 2” CL bit. I did increase the dimensions of stationary steel beams to 8”x 3”x 1/4” (from 8x2) and the fabricated aluminum traveling beam to same dimension. I figure all in with all components and hardware, the traveling gantry with Z axis and router will weigh about 160lbs.

I’ve been revisiting my thinking on motors and drives. Just checking myself here but:

As a recap, my top (foam) machining speed is 200in/min (5080mm/min). I have designed for 10mm lead ball screws so the ball screw rpm would be 508 rpm direct drive. I’ve got a 3:1 timing belt mechanism designed in between the stepper and ball screw as shown in the attached, with the intent of keeping the stepper in a more optimal torque range.

Looking at the attached torque curve, it doesn’t look like I’m really gaining anything by adding a timing belt. With a 3:1 pulley ratio the torque at the screw is reduced to 1/3 that at the motor shaft. Looking at the torque curve of the stepper, the stepper torque looks to be about 1/3 at 500 rpm (maybe only slightly less), so considering torque alone, it seems adding a timing belt to reduce stepper rpm does not offer much if any benefit purely in regard to torque delivered to the lead screw and it’s just more complexity and opportunity for position error.

Stepping resolution seems to work for better in direct drive as there could be three times as many steps for a given stepping setting per screw revolution. I don’t gain or lose anything wrt to screw whip, wear, or load, with either because the screw speed is the same in both cases.
 
I would only be running these speeds cutting foam, and frankly, most of the time even that would be in the 100-150 in/min range.
Now my existing machine rapids at 650in/min. I’m not sure that’s a possibility for the new machine even though it would have two steppers and screws driving it, but I’m not inclined to value rapids speed highly because in my mind, if the machine is moving it should be cutting, not moving to a new position, at least the vast majority of the time. I figure that will be whatever max it can be.

So I’m thinking I may as well direct drive the screws with the steppers, but still may require a timing belt on the moving beam just for packaging to allow the beam to be repositionable. Am I thinking right here? Anything else I should consider in this?

I was a bit shocked by what decent quality timing belts and pulleys cost. It was going to be $150/axis just for a set of timing pulleys and belt. I could almost buy a gear box for that, and approaching the cost of import servos instead of closed loop steppers if they could be used with direct drive. -I know squat about servos. 

Best,
Kelly

[dave benson]

HI Kelly

With machine design, “less is more” If at all possible avoid using the pulleys, as you’ll save
the initial installation cost and ongoing maintenance as well as the weight savings on the gantry
which for a moving gantry is a big deal.
I was watching this video the other day, How NOT to Design a 5 Axis CNC - YouTube
I was mainly interested in the 5 axis code, anyway the young man did two videos one of the build and a follow up video (piece to camera)
about what would need to be changed in a future build. A lot of engineers posted in the comments saying that his observations were correct.

A servo drive has it’s rated torque up to it’s rated speed and doesn’t drop off like a stepper.
So for example if you replace the X axis stepper with a 3000 rpm servo, and gear it down with the pulleys to your required 500 rpm
then you get a 6x increase in torque and resolution meaning that you can select a relatively small and therefor cheaper\lighter servomotor.
The only downside of a servo is the price.

The hobby servos that you would be interested in have step and direction inputs, meaning that they
are a drop in replacement for steppers, be careful of the industrial servo’s as they might not have
that interface as it’s to slow, for example with a Clearpath servo you connect up the power and step\dir inputs to your existing controller and off you go.

With the gantry endplates, there are a lot of videos on utube were the Y axis rails are mounted
higher up off the bed to keep the endplates short and stiff, this also drastically cuts the moving mass
of the gantry.

[Tool-n-Around]

…..I was watching this video the other day, How NOT to Design a 5 Axis CNC - YouTube

Funny, I just watched hat (Mach Super’s) series last night.

A lot of engineers posted in the comments saying that his observations were correct.

I did take note of those comments. I may increase the fabricated aluminum X-Beam from 8x3 to 8x4 just for a little more torsional rigidity. Problem is, the spacing on the linear bearings on the gantry plate eats up a lot of Y-axis travel, in my case, 12”. With other packaging, it means I have just over 625mm usable of my available 1000mm linear rail . Even with my previous efforts to minimize the distance to Z/spindle centerline, along with the thickness of the X-Beam, I still can’t get my spindle axis centered between those two bearings on the gantry plates, but I have 25mm rails and bearings on the Y so I think I’ll be just fine as far as bearing loading.

…….A servo drive has its rated torque up to its rated speed and doesn’t drop off like a stepper. So for example if you replace the X axis stepper with a 3000 rpm servo, and gear it down with the pulleys to your required 500 rpm then you get a 6x increase in torque and resolution meaning that you can select a relatively small and therefor cheaper\lighter servomotor. The only downside of a servo is the price.

Stepping down 6x….hmmm. I had not considered that. I suppose that would also mean any small error introduced by the belt and timing gears would be 6x smaller too. Those long stack NEMA 34s are big and weigh 5.5kg. Taking a quick look at the Steppers Online offering, their E-series only weighs 2.5kg and is quite compact. With driver they are $230/axis with driver plus $150 worth of timing belt and pulleys. With 6:1 reduction it would be $380/axis…..or $1520. More for their higher end T series.

The E-series 1000w servo is 3Nm and $260/axis, which is about the same torque as the closed loop N34 stepper ($160/axis) at 500rpm with no reduction, but the stepper torque triples at the lower speed range which in my case is the majority of use.
Being the machines intended use and being a hobby machine, I sort of doubt that I could really fully appreciate the benefits of servo vs N34 driven system, especially if I can direct drive the steppers. The way I have it laid out, the hardware could fairly easily be retrofitted with either a step up or down timing pulley system if I got into a jam.

The hobby servos that you would be interested in have step and direction inputs, meaning that they are a drop in replacement for steppers, be careful of the industrial servo’s as they might not have that interface as it’s to slow, for example with a Clearpath servo you connect up the power and step\dir inputs to your existing controller and off you go.

I’m not quite at the point of pulling the trigger yet but for my relative level of cnc sophistication and use, it’s probably silly to do more CL N34s. Either way it’ll be a huge step up for me.

…With the gantry endplates, there are a lot of videos on utube were the Y axis rails are mounted higher up off the bed to keep the endplates short and stiff, this also drastically cuts the moving mass of the gantry.

I sort of split the baby on that. My bed surface is halfway down the height of the 8” tall steel Y Beams so with the bottom of traveling beam in lowest position, the traveling X beam is at the very bottom of the gantry plate. It will probably be rare that it sees duty in the more prone/higher positions.

Thanks for all the input Dave. Much appreciated.

Best,
Kelly

[Tool-n-Around]

I don't spend as much time in the shop during the fair weather seasons but still plugging away. Here's the gantry beam. It's an assembly of the two lost foam castings end to end, and then the two faces get skinned with 1/4" aluminum tooling plate. Each casting half weighs 14lbs and together measure 44"x 8"x 4". Everything gets screwed and glued together with structural adhesive.

I didn't want to make them in halves but it's sort of with my capabilities as far as casting and rough machning for assembly. Even so, the two finish machining operations on the ends and prepping the surface for the rails will need to be done at my buddies shop because I dont have the travel nor accuracy to machine the surface for the linear rails. I figure it will weigh 40% and be at least as stiff as a 8"x4"x1/4"w rectangular steel beam. The gantry plates will also be castings but the rest of the structural rectangular steel.

https://youtu.be/DOpBRa-_xT0
https://youtu.be/j4okXkw72eM

Best,
Kelly

[dave benson]

Nice video’s Kelly, I was wondering how you were getting on a few day’s ago
as you keep up a frenetic pace and hadn't posted for a while.
A forty percent weight saving for the same stiffness is a good thing and well worth the
time and effort.
They are big castings, looking forward to see what you do with the gantry
side plates.