[crjohnson]

G'day guys,

I managed to get my hands on a pair of CSF-32-50-SP harmonic bearings and so have started designing a 4th axis for my cnc.

I have been looking into tapered roller bearings for the spindle and balancing up availability, size and cost. My original design had 60x85x17 size bearings but I just found some 65x100x23 SKF units for a reasonable price.

These are quite a bit larger and heavier so I was wondering if anyone has experienced a scenario where bigger wasn't better? I don't believe their inertia will be an issue due to the stepper I am looking to use. But wasn't sure if getting the right amount of preload would become difficult or anything else.

I plan on mounting a 100mm 3-jaw chuck or a PAF-H54 style vice and also a NEMA34, 4.5Nm stepper with a belt drive.
Cheers!

[crjohnson]

Here is a section view for reference. Still a few details to finalise (shoulders for the bearings, spacers etc.).

[peteeng]

Hi CR - I suggest you look at automotive unit bearings used for wheel bearings. They are in either ball or tapered arrangement in a single outer race. Being preset in the factory means you just need a nut to pull up the inner races and they are done. They are available in a variety of sizes from small to very large. But they are not sold via size so you have to work through the car models and figure out which one you want. I have a suzuki swift bearing and a medium sized SUV bearing here for my next rotary axis and they are very good... Peter

[crjohnson]

Hi Peter,
they were actually my first port of call for looking at bearings. Then I found some large tappered bearings (originally for wheels) went down that route..

I will revisit that idea though as being preset may simplfy things for me considerably!
Cheers

[crjohnson]

Here are a few options. Looks like most are in the 40x80x40 size range.

The toyota listings were from a 2018 hilux, suzuki were a 2015 swift, hyundai was a IX35.

[peteeng]

Hi CR - Here's one I use on one of my trailers $15000 ID 416mm Peter

[joeavaerage]

Hi,
beware that automotive wheel bearings will often retain a 'little bit' of clearance even when done up snug. Wheel bearings are under high load at highspeed and therefore a little bit of bearing clearance is
often required to offset thermal growth. A CNC axis will never generate that sort of heat and therefore the bearing may retain that bit of clearance. I was a mechanic for twenty years and even when unitised wheel bearings
were torqued to manufacturers spec, many hundreds of Nm commonly, there is still clearance.

Craig

[peteeng]

Hi all - If you want to get technical with unit hubs here's some design info. There is a rigidity calc in here, plus mounting info etc. Bit of a mouthful but could answer some questions if you want to go that deep. Peter

[crjohnson]

Thanks for the replies guys, I am leaning towards the large tapered roller bearings due to dimensional constraints. The wheel bearings are too small and cause issues with attaching the output of the harmonic drive.

I also guess the 4.5Nm stepper will be more than capable of handing any inertial loads I put on the system.

[joeavaerage]

Hi,
I have some months ago completed a fourth and fifth axis. I make a couple of points that you might like to ponder.

Firstly I used worm type servo reducers rather than harmonic drives, but the same comment applies. Namely that there is significant frictional drag involved in turning the input shaft
irrespective of the torque of the load. In my case the reducer is 19.5:1 (fourth axis) and 6.75:1 (fifth axis) while I imagine your harmonic drives will be at least 50:1. You might well imagine
that any likely load will still represent a small load to the stepper....but not so the frictional load, that applies even with no load.

Initially I used a 4.5Nm stepper for my fourth axis but found that the frictional load marginally overloaded the stepper, and it would occasionally miss steps. I then replaced the stepper with a 750W
servo, much faster and never missed a beat since.

Second issue is resolution. Lets imagine your rotary axis has a radius of action of only 100mm. If you wish to have a 0.01mm resolution at that 100mm radius that requires an angular resolution of
0.1 milli-radians or 20.66 arc seconds. This came as a bit of a surprise to me, it should not have, its plain trigonometery and algebra after all, but none the less it did.

The bottom line is that even with a 50:1 reduction will still require a fine resolution of your stepper, indeed you may find that the resolution somewhat beyond a stepper, indicating a servo instead.

Given say my fifth axis reduction of 6.75:1, the required resolution to achieve a 20arc second output resolution is quite demanding, rather more so than your 50:1 reduction. I used another 750W Delta servo for the fifth axis platter.
Both speed and resolution....whats not to like?

May I suggest you do the resolution calculation and establish whether a stepper will be adequate.

Craig

[peteeng]

Hi CR - I understand the issue. You will need to preload the bearing for stiffness and impact survival.
Hi All - the unit bearings come preloaded from zero to some preload (mentioned in the NSK doc). If you run a bearing like this (tapered or ball) in a situation that has impact with clearance they will Brinell and fail. This is the same reason that linear carriages are supplied preloaded for stiffness and for impact survival in cnc machine use. Excessive preload shortens their life but a correctly sized or prefer oversized bearing will provide adequate stiffness and life with quite a high preload. Since auto wheels get quite hot they expand and their initial preload decreases so the factory set preload must be well above the service preload. So I expect since on a cnc they won't get very hot the unit would stay in a preloaded condition. I'll have to set up the bearings I have and check that that when I get there on my rotary table design.

Looking fwd to your design details. Peter

[joeavaerage]

Hi peteeng,
I did these things for twenty years, most unitised automotive bearings have clearance....that is if you like their 'preloaded condition'. No matter how tight you do them up you'll
never remove that clearance, not the best for CNC.

That clearance is absorbed as the bearings heat up, so they have little or no clearance at operating temperature. High speed bearings need MORE clearance....not less.

Craig

[samco]

I used a wheel bearing for a 4th axis. (it was like a toyota or something)

As joeaverage mentioned - they might not have pre-load when clamped together. I found this out. I ended up popping one of the inner races out and grinding a few thousands off of it until I had the pre-load I wanted...

I talk about it here...

[joeavaerage]

Hi,

I ended up popping one of the inner races out and grinding a few thousands off

Good idea.

What surprised me often is how much clearance Subaru specified for their unitized wheel bearings. Subaru's are for some reason hard on them and we had to do a lot of them over the years. Often you'd put a new bearing in,
torque the axle nut up to 230Nm, or whatever was specified, put the wheel back on only to find there was 1/8th inch side-to-side play at the periphery of the tyre. You'd swear the bearing you'd just put in was faulty. Might have been
fine for a Subaru wheel bearing but would be hopeless as a CNC axis bearing!

Craig

[crjohnson]

Joe: here are some calcs I had previously done and compared them to some commercial indexers/ 4th axis units.

I was looking at using a 1:1 ratio belt drive between the stepper and harmonic as I didn't want to make the axis too slow.

I thought that if my unit looked comparable to them, then it should do whatever need! Hopefully my calcs were correct too.

[joeavaerage]

Hi,
I think your calculation of stepper resolution is so wildly optimistic as to be useless.

You may be tempted to think that if you assign a 3200 step per rev that the stepper has a resolution of 360/3200=.1125 degree or 6.75 arc min. Unfortunately, you'd be wrong, well wrong.
You can assign any micro stepping regime you like but the practical stepper resolution is no more than 400 steps/rev or 0.9 degrees or 54 arc min. When that is translated through
the harmonic drive that equates to 64.8 arc sec, well shy of say 20 arc sec which should be your target.

The remainder of the calculation is reasonable, if perhaps a little optimistic with regards the output torque, but still a good ball park figure

There are a few posts about what resolution you can expect from a stepper including the phenomenon I have described, but it comes down to the torque between micro steps diminishes as the number
of mirosteps increase.

Craig

[joeavaerage]

Hi,
this link explains it in more detail and more authoritatively than I:

https://www.monolithicpower.com/why-...0full%20steps.

Craig

[crjohnson]

haha ok, I picked that microstepping figure as that is what my X-Y axis run on the machine. Z axis runs 2000steps/rev.

I'll read over your link and see if a reduction ratio on the belt is needed. Cheers

[joeavaerage]

Hi,
My steppers were five phase (Vexta), so 500 steps per rev, but also they were through low lash 10:1 planetaries, and consequently I could guarantee 5000 steps/revolution or 4.32arc min per step.
So I had the resolution but it came at the expense of speed.

Many of my fourth axis toolpaths were continuous rotation through many hundreds, sometimes even thousands of turns. They would take an age to run at just 10 rpm (max fourth axis output speed).
That is why I went for a servo, both I wanted and needed the resolution, but also wanted a much higher speed that my toolpaths would run quicker. With a 3000rpm servo the fourth axis now runs at 150rpm
so, a better than ten-fold increase in speed, and still get nearly four times the resolution. Win,win. But it cost a sevo and drive, $438USD plus shipping for a 750W Delta B2.

Craig

[crjohnson]

Thanks Joe, I'll have a look into a servo drive.
Even at 1000steps/rev (at the stepper motor), I end up with 0.0072deg/step resolution and 0.012mm at 100mm radius. Maybe 3200steps was overkill but I started there as thats what the rest of my CNC runs and it seems to work good. Maybe I can back the CNC motors off a bit to say 1000-2000? 5mm pitch ballscrews on every axis..