Nomad with deep DOC, shallow WOC

CNC Machines Nomad
161

Yes, I considered Sorotec’s various offerings, including the Teknomotor products. Two categories I ruled out entirely:

  • HSD and Elte had a worrying lack of specs (even basic things like runout)
  • Teknomotor’s “HF Motor” spindles only have 18k RPM and claim low tolerance for axial loads

That left the Teknomotor electrospindles. Those looked good and I like that they have electronic fans but I noticed that they’re larger and heavier than the Mechatron spindle (4kg for 800W vs. 2.4kg for Mechatron) and when I searched around, I couldn’t really find anyone who’s used one.

I did consider the ER-20 collets. I’ve seen a bunch of my machinist YouTubers using indexable tooling and thought that surely I should have it too but when I read up on it, it didn’t seem like a clear win. The Nomad is a small machine and I don’t think I can physically fit in pieces of stock large enough to use such large tools.

As for larger endmills, that’s true but most of the endmills I’ve seen have recommended surface speeds around 500m/min, which I already get with 6mm tools at 30k RPM. And again, bigger tools work great for bigger pieces but the Nomad can’t physically fit a whole lot in there.

And I know I sound obsessed when I talk about MRR but I don’t really need anything extreme. The Nomad currently maxes out at ~0.5cm³/min before it hits its limits, which is… Insufficient… But with the new spindle, the 30k RPM alone allows me to get ~9x MRR over the stock spindle, while keeping the same cutting forces. That means a job that previously took the whole day takes just an hour now.

With a Nomad that’s no longer like a blob of jelly and the improved spindle, I might be dreaming, but it could be possible to push it to 42cm³/min of MRR (84x stock). At that point, the modded Nomad can do in an hour what takes a stock Nomad 2 weeks. That comes with peak forces of ~82N but with axes built on ballscrews and linear rails, maybe it’ll be okay?

To be clear, even if it’s possible, I don’t think I’ll routinely push the Nomad that hard, I just wanted to think about what might be theoretically possible.

And back to ER-20, I won’t have it now but since I have flexibility, I can always swap in a new spindle in the future, just to try it out.

162

I’ve also been tempted by the idea of some insert tooling but Luke’s very helpful post here damped that enthusiasm quite a bit;

1 Like
163

Wouldn’t it be 3X MRR (3X Speed/Feed)?

164

When I put it into Millalyzer, it showed ~1/3rd of the force on the spindle even with the tripled feed rate, so I could bump up the depth of cut 3x as well while keeping the same force.

165

Maybe you forgot to triple the feed rate?

1 Like
166

Ah, sorry, you’re right. Usually I have Millalyzer on feed per tooth mode and it does it for me but apparently I didn’t when I did this.

1 Like
167

Can you actually use a 5/16 for 8mm??? Its darn close… never even thought of that!

168

I haven’t had a problem using them interchangeably — I mean, I’d like to see the production line which reliably holds specs where 5/16" == 7.9375 millimeters is significantly differentiable from 8mm — scarcely 2 and a half thousandths.

2 Likes
169

That’s less than the size and runout tolerance on many of the collets.

2 Likes
170

So I realized that there’s a major deficiency in my current design: It lacks supports for the ballscrew.

This isn’t a problem for the ballscrew itself but at the moment, the design puts the entire weight of the spindle and all the axial load and vibrations directly onto the stepper motor shaft. I think this is a bad idea. It was fine for the regular Nomad but the regular Nomad’s Z-axis carriage doesn’t weigh very much and the spindle isn’t made for axial load anyway.

I was also looking through the datasheets for the rails and I think there’s another problem: the required vertical parallelism of the surfaces the two rails sit on (the allowed height difference between rail 1 and rail 2).

The TBI datasheet says I get 120µm tolerance for Z0-preloaded blocks but their image also shows a 500mm separation between the rails. When I look at the Hiwin and CPC datasheets, they calculate this using an equation of the form tolerance = <constant> * separation, where tolerance is 4e-4 for CPC and ~2.6e-4 for Hiwin. If I assume TBI rails are governed by a similar equation, the constant would be 0.120/500 = 2.4e-4. With my current 88mm separation, I get:

Brand Constant Tolerance
CPC 4e-4 35µm
Hiwin 2.6e-4 23µm
TBI 2.4e-4 21µm

And to add to the complexity a bit more, there’s variance in the heights of the carriages themselves. For the standard-precision carriages, the carriage height tolerance is usually ~100µm, “high” and “precision” are ~30µm, “super” precision ~15µm and “ultra” ~8µm.

Theoretically, those numbers should be subtracted from the flatness tolerances above, so I even if I go for the “ultra” precision carriages (which cost ~2x), I still need ~27µm flatness for the mounting surface.

I could try machining this on the Nomad but I fear trying to get that level of precision out of the Nomad would drive me to insanity.

@spargeltarzan, do you have any clue how expensive it would be to get a plate with those tolerances machined at Spanflug or Weerg? I’d ask but I’d have to ask them for a proper quote and not just use the automated tool.

Or is there somewhere I can just buy mounting plates made for this?

Or am I overthinking things and it’s totally fine to just ignore the specs and throw the rails on a faced plate of Aluminium?

171

Shim it to win it buddy

172

Hmm, not a bad idea. It looks like I can buy rolls of 10µm steel pretty easily and by stacking them I should be able to get the rail to within tolerance.

Is there a way to do it without a surface plate though? I’ve been putting off buying one because they’re so heavy and expensive.

173

Are you planning to use the flange or block type block?
“There’re two types of blocks:flange and square. The flange type is suitable for heavy moment load application because of the lower assembly height and wider mounting surface”

Where did you find the equation?

I was initially confused by the HiWin docs and mistook the parallelism between rails of the reference edges § for the horizontal surface flatness (S1).
For ZA preload class HGR15 this seems to be 85um.

For the rail parallelism, I plan to follow the common YouTube user pattern of bolting and attempting to straighten the “primary rail”, whose blocks sit against a “reference edge” on my carriage plate. I will then use the blocks on the carriage plate to align the “secondary rail” sufficiently parallel to the “primary”.

As Vince says, shimming under the rails is also quite popular if there are large deviations to fix.

Also, for @Julien on pg 28 of the HiWin linear rails doc;

Screenshot 2020-11-15 at 17.52.57
Screenshot 2020-11-15 at 17.52.571724×536 58.9 KB

1 Like
174

Block type. Flange type is too wide given the limited space available.

In the Hiwin Linear Guideways catalogue from here, at the end of each section. For example section 3.1.14.3 on page 45. Strangely, I checked the American version and they don’t mention it they just give the values for 500mm separation.

Also in the CPC catalogue, page 25 (printed) or 16 (PDF page number).

And THK also mentions that the number is “proportional” to the separation, which I think is saying the same thing, on page 1-453 of their “Designing a Mounting Surface” PDF.

No, I can see how that could happen but I’m quite sure I’m looking at the numbers for vertical parallelism. I’m not too worried about horizontal, as long as there’s a bit of wiggle room, there’s plenty of ways to deal with that.

1 Like
175

That’s really interesting, maybe they think that only Germans are detail oriented enough to use the formula and not just look up a number?

If you look at the geometry of what’s going on and the specs for the rail block (also the symbols they use in the drawings) it seems that what is critical is the ‘flatness’ of the mounting surface. Specifically, the rails must be normal to the surface to a high precision as any rotation of the rail is amplified across the block to block mounting distance.

Having watched a few folks mounting HiWin class rails on DIY CNC on YouTube, people seem to have been successful mounting these rails onto surfaces which are quite achievably flat for our purposes. I’ve also test-mounted mine here with Z0 and ZA preloads. There’s lots of people out there bolting rails to extrusions, tooling plate and poured epoxy surfaces and getting away with it. If you look at mods like @DanStory apparently the face of the Carbide extrusions is flat and parallel enough.

We may not achieve the wear lifetime and some other elements with poor alignment but it will likely still work. The other thing to consider is that HiWin is telling you that they warrant the rails and blocks to operate within spec over a full lifetime at the stated loads if used in an assembly meeting the specs. The use of reference edges and flat, parallel surfaces means that when assembled in a factory there is a good chance of things actually working.

If you have a major misalignment what you will likely find is that the carriage becomes very stiff or binds up completely at the point of misalignment, strategic loosening of bolts reveals where and allows you to scrape, shim, push or otherwise sort that alignment. (My HDZ had a ball nut whose mounting face was not parallel to the axis of the ballscrew on delivery and that was reasonably quickly identified by different combinations of loose & tight)

2 Likes
176

Though I find the idea amusing, THK and TBI do the same thing as well so I don’t think that’s it. My guess would be that a German wrote the German version of Hiwin’s documentation or a German complained to Hiwin and got the German documentation upgraded.

I think the problem might be that when the surfaces aren’t the same height, you’re basically preloading the bearings with a moment arm? Linear rails aren’t good with radial loads, so this would naturally shorten their life.

I did a bit more research and epoxy also came up for me as a recommendation but I think I’d prefer shims to epoxy.

I actually came across this as well and started to think about lapping… I can get 3 pre-milled 7075 plates cheap enough and apparently water and Aluminium oxide works as lapping fluid. There are also other options like dual disk grinding. I might be overcomplicating things though.

I think I should just bite the bullet and finally buy a surface plate. No matter what solution I end up with, I’ll want the ability to test it.

Also, just briefly on the Hiwin catalogue, holy moly that’s some serious stiffness. The bog-standard HG15C has 200N/µm stiffness. Just one bearing would be enough for the Nomad.

1 Like
177

hmm there are even 3/8th collets for ER11

178

It’s almost like you Germans want to do things properly so they work, not just sneak off to the pub at 16:45 or something :wink:

Lots of people have fitted these rails to less precise surfaces, I’d buy them, make the major parts, fit them and then see if the flatness is really an issue.

Since deciding Xometry weren’t going to deliver anything worthwhile (another 2 weeks of no response from their ‘customer service’ team now) I discovered that I could get the adapter plate flatter than I can measure on the Shapeoko with some tramming followed by post-polishing with fine grade wet paper on a big piece of float glass and then a scotchbrite pad.

Indeed, these things are rated for silly loads, I’ve tried to measure deflection on the Carbide HDZ and failed, I can’t find a reasonable way to put enough force on it to actually measure any deflection and that uses more ‘commodity’ linear rails from what I can tell.

179

Actually, the spanflug automatic quote includes the cost of any additional accuracy requirement that is properly specified in the drawing. Last time I had a quote with 0.02 mm concentricity requirements, the price went up by a few euros per piece.

That said, I suspect that it should be possible to get close enough with the Nomad if it can be done in a single setup (that I don’t know). As @LiamN pointed out, you are probably very far away from the specified limits (load and linear speed) to worry much about lifetime. Ingested dust or total lack of lubrication would be more detrimental in that respect.

3 Likes
180

Well I ordered a few hundred francs of metrology stuff and shims…

I think I’m mostly fine on the design now, everything is parameterized and waiting for whatever ballscrews and rails I put in. I’m waiting on quotes from various folks for those. Tuli is okay but it’ll take them 3 weeks to machine the screws, which is… Not what I’d like.