Why not use unsupported rails?

This question arises in almost every build log on the forum and every time someone explains why they are unsuitable for most CNC Routers. Here are some simulations I have run to prove the point...

All rails are 20mm diameter with 150mm between bearing block outer faces. The bearings are attached to a 75x160x20mm plate to which the forces are applied.

Assumptions:
Rail / box section ends are perfectly rigidly supported - in reality wont be so deflection will be greater than calculated.
All joints are perfect, modelled as single entity - in reality will get some deflection between joints.

The following diagrams each have different scales to make the deflection visible throughout.

Unsupported rail 200N in Z direction to simulate weight of Z-axis:

Attachment 5615

Supported rail 200N in Z direction to simulate weight of Z-axis:

Attachment 5612

Unsupported rail 200N in Z direction to simulate weight of Z-axisand 50N in parallel to X for cutting force:

Attachment 5614

Supported rail 200N in Z direction to simulate weight of Z-axis and 50N in parallel to X for cutting force:

Attachment 5613

Max deflection for supported 0.0074mm vs 0.34mm for unsupported... draw your own conclusion.

Bingo! Thanks for that Jonathan
Bruce

So by the time you have factored in unsupported rails, rolled ball screws, or Acme screws, off the shelf commercial extrusion, lovejaw couplings, roller skate bearings as thrusts for ballscrews you have the equivalent of a high tensile wet rice krispies box.

Good job the builders of these machines don't have access to a FEMA program so they can see where they have gone wrong. :rofl:

So, for those of us who are colour blind:

Case 1) Unsupported deflection is about 100 times greater than supported
Case 2) Unsupported deflection is about 50 times greater than supported

Nice work J. Ever thought of designing roller-coasters?

This thread just got sticky :tup:

I think they would make a good set of handle bars.

Very pretty but also a very non accurate picture has been painted here. I use 30mm un supported rails on my machine and the accuracy is as good as can be (way more than my needs). The main factor is that it is all bolted to a very sturdy structure. the problem with your assumption here is that the rails sit on something with a solid base. The main problem with machines is the deflection of the bed and not necessarily the rails? if you look at all those nice machines with expensive supported rails or slides, most of them are mounted on aluminium profile that will flex more than the rail. You need to recalculate this to show the defection of the bed for it to stand up? Also what is this for Milling or routing?

Thats my 2 pence worth

Quote:

---

Originally Posted by 2e0poz

Very pretty but also a very non accurate picture has been painted here...

---

You seem to be under the impression that I'm trying to say that unsupported rails are weak. I'm not, I'm saying they are many times weaker than their supported equivalent. If 30mm unsupported rails give you enough accuracy, then I'm saying you could achieve the same result on a much smaller diameter with support.

If I'm honest, I've seen so much abysmal design in commercial machines that I simply ignore what 'expensive' machines do. Their objective is to make money not make good machines, and those different objectives result in different designs.

Yes, there is deflection in the bed, I'm not implying this accounts for all the deflection. My only point here is that the same diameter of rail is tens of times weaker when unsupported, so all other factors being equal, if given a choice between unsupported and supported, you should go for supported.

An FEA simulation is never 100% accurate, but it's certainly a big improvement on estimation and common sense. I've simulated bed designs too, here's park of one I'm working on:

Attachment 5618

Jonathan i make no accusations, i just responded to your weak bold statement that lacked accountability of all facts. It was you that made the statement that unsupported rails are weak not me. I just pointed out some of the missing facts? If you could show how strong supported rails are attached to 40 x 40 profile like a lot of machines are you will see just how weak they are in comparison to your model?

PS I'm a performance engineer by trade and do a lot of modeling. i would get shot down in flames if i presented statements like that without all the facts.

Quote:

---

Originally Posted by 2e0poz

If you could show how strong supported rails are attached to 40 x 40 profile like a lot of machines are you will see just how weak they are in comparison to your model?

---

And your point is? They'll still be stronger than the same size of unsupported rail which was the point Jonathan was making.

Quote:

---

PS I'm a performance engineer by trade and do a lot of modeling. i would get shot down in flames if i presented statements like that without all the facts.

---

Obviously not a great one since your using unsupported rails :joker:

Use some common sense in this thread, realise this isn't a white paper on rail strength and that everyone here is a hobbyist

Our bouncy friend is correct in saying that we should also consider the worst case examples for supported rail.

1) Supported rail rigidly supported at both ends but without any underlying box section.
2) The deflection of the underlying box section sans rail.

Why?

Well, 1) gives us some idea of the contribution of the basic rail support structure and 2) gives us some idea of the contribution made by the box section itself.


Edit:
I would love to see a similar analysis of profile rail. Just how significant is the support structure? 20mm plate, 40mm and 90mm section, aluminium and steel. After all, £600 worth of rail is meaningless if the £100 worth of extrusion allows deflection on a par with round rail.

Quote:

---

Obviously not a great one since your using unsupported rails :joker:

---

I bought my rails a good few years ago now before supported rail became affordable to the DIY scene. It is too easy to put what is a practical use down because of listening to misinformed information. From a reality point of view i have not seen anything like what Jonathan is insinuating about using unsupported rail. It is also putting people off even trying to build an affordable machine that may well have been suitable for them. A few people have spoken to me asking for a realistic picture of using round rail because they had access to it. They have been worried that they may get shot down for doing something bad.

now if you were going to span a 20mm rail over 1.2m then you are asking for trouble. It is all about using the right thing for the right job so comparing the same size of rails against each other is pointless.

As for my job i do system engineering not mechanical but my rail was spec'd by an mech engineer which is why it works.

Well we had better start looking on Ebay for some real cheap routers that only need upgrading to supported rails.
That will cover most of the Techno Isel range, Hiez, and all the Brit ones.

I'll grant you supported rails will be far stronger than unsupported, it doesn't take a white paper to work that out [ unless it's toilet paper - non coloured ] but how about doing a paper on what is allowable in a DIY enviroment.

Has any one else ever mapped the deviation on say an X rail as it travels full length due to bed twist, and then done the some on the other rail as it returns back?

Those results will be very interesting. A quick and dirty test on any machine is to do a very fine engraving in brass then run the code again and see what extra metal is removed on the second pas or if it's thinned the engraving down. This is a very good PRACTICAL test.

John S.

Your right john, it would be interesting to see what the equivalent would be of unsupported against...for example would 50mm unsupported be the same as 25mm supported?

The other thing to consider in all this is that one uses open bearings and one closed, what deflection do you get off those?

A lot of the British machines are using designs from eons ago with very little change which is a shame as the ones from afar seam to at least try new things. Now Mary Portas has the right idea of getting Britain manufacturing again. I'm sure we would we streets ahead if only we got our hands dirty. But for now we will have to stick with pants over here....

There are so meny variables when building a machine to it's accurecy. All this talk of unsupported or supported, profiled etc is a bit lost and pointless on 95% of DIY builds because of other areas of the machine that are completely overlooked or under estimated.!! . . .The BED and Z Axis being two of the main one's.

Most on here will Know by now I'm not into Calc's or simulated predictions so my view is based purely on personal experience and for machines over a certain length or width I just would not recommend or use unsupported rails. This Max length and width would be 300mm for X & Y Axis and I would NEVER use them on a Z Axis.
I've only ever made one Z axis this way and it was by far the weakest area of the machine and big regret, but alas and as is often the case it was down to cost saving as usual.!! This is why I URGE any one building to just spend the few extra quid and save them self's this problem and go for supported rails.

Profiled rails are very misunderstood.?? You don't build a machine with profiled rails for there strength, infact they would probably show a tad weaker than supported round if simulated like jonathan did.?
You use profiled rails for there accurecy, smoothness as well as there abilty to handle far far higher loads and in DIY usage for millions of years.!!
To be honest you don't have to be an engineer or an expert to see and feel the difference between round rail and profiled, the quality just ouse's out and by comparison round rail, supported or unsupported, are like baggy fanny's in comparison.!!

Another often unknown thing about profiled rails is there intolerence of poor design, shoddy materials or poor workmanship. They will bind and lock with the slightest of miss aliagnment or uneven surface so this again is the test of a well built machine.

John S test is a great test and if you want to test the accurecy and more so the repeatabilty then cut the Aztec calendar because this will soon show you how well your machine is performing. The Aztec calendar as very fine detail and with 855,000 lines of code It boucnce's about like a pogo stick so if your machine is out in the slightest degree it will lose position and show any error when recutting so removing any detail.

The purpose of this simulation is purely to compare rails, not to say one type of rail is better than another in all circumstances, clearly each of the 3 main types of rail has its place.
What the supported rails are mounted to has to be included, as nobody would mount them without any support, however including other factors such a bed strength is unnecessary complication and so should be considered separately. Clearly when designing a real machine you must consider how all the parts interact, so ideally one would simulate the machine as a whole. However since the rails are a major source of error it is reasonable to look at them independently to help with selection.

The main advantage with profile rails is they can tolerate high preload, so the stiffness is much higher and there is no play (so long as the preload is not exceeded), which is why I will soon replace my round rails. Round rails have much less, if any, preload so the deflection of the bearing block relative to the rail is non linear and greater. In the case of just supported rails the load rating/stiffness varies depending on the direction of the force which is a significant disadvantage if not mounted 'opposing'. The accuracy of a profile rail is not much different to a round rail - it depends how you mount them as how many of us correct the bend in a profile rail by comparing to an accurate straight edge reference? Not many as for the vast majority of machines it's excessive.

Quote:

---

Originally Posted by 2e0poz

Your right john, it would be interesting to see what the equivalent would be of unsupported against...for example would 50mm unsupported be the same as 25mm supported?

---

Depends what you mount them to...I've run simulations to find the equivalent in supported rail of 30mm unsupported. Since you suggested comparing 40mm extrusion I have used 40mm *light* aluminium extrusion. Same forces as before.

Table, Ratio colum shows deflection relative to the equivalent loading on 30mm unsupported.

Attachment 5619

Pretty pictures attached:

Attachment 5620

So the 30mm unsupported rail is similar in bending to the SBR16 rail on weak 40mm extrusion, yet currently using the cheapest prices for each I can find the 30mm system costs 70% more. That's for the rails, bearings and end supports, or supported rails, bearings and extrusion. That implies that now the prices of SBR (and TBR) rails have decreased significantly since 2eopoz bought his it is no longer makes economic sense to use unsupported. The exception of course is a small machine (Jazz's 300mm estimate sounds reasonable) with lower forces, or laser and plasma cutters.

The thing to do now would be to run the simulation with two rails as that is more realistic, however I don't think it will tell us anything new.

Quote:

---

Originally Posted by Jonathan

The accuracy of a profile rail is not much different to a round rail -

---

Sorry Jonathan that's Rubbish total utter rubbish they don't compare in any way.!

Just the fact the bearing blocks are ground flat and true on profile bearings makes a massive difference. Bearing in mind 90% of round rail is cheap chineses tuff I've yet to see a round rail bearing block that's flat on top with sides machined 90deg. Only the very best brand quality are any where near and these can cost the same as profiled.

Then we get to the actual bearing it's self, the round rail bearings are sloppy from the begining just use a dial gauge and apply side pressure to see that, you won't find this on profiled bearings due to preloading and the precision machining of the rail.

They are night and day apart.!!

Quote:

---

Originally Posted by JAZZCNC

Sorry Jonathan that's Rubbish total utter rubbish they don't compare in any way.!

---

Sorry I should have been clearer, I meant the accuracy of the straightness of the rail, not the tolerances they are machined to. Obviously I agree that overall profile rails are machined to much better tolerances and thus by far the most accurate rails. Round rail bearing blocks and supports are merely extruded.Sloppy - i.e. clearance not preload which is exactly why I'm replacing mine. Simulation of a realistic gantry, same as in first post except both rails and 10mm aluminium plate stuck on the back, forces in X and Z and constrained at the ends:

Attachment 5622

Actually Jonathan you have now shown a fairer account of the simulations, this shows a more realistic picture so well done to for that. The deflections shown though in the grand scheme of things show very little in the way of movement between them all looking at those figures. Now if i were to build again i would 100% go for linear rails supported on the biggest bases as possible, in a small desktop machine for pcb and plastic work then round rail would be fine (only if it was the cheapest and only option).

Thing is a lot of people expect their routers to be milling machines as well, only in this case you would need to worry about expensive linear movement. For most diyers the need to be honest about what they want to use there machine for and build accordingly. If they can not afford that then they have to live with the compromise or save up and wait a bit longer.

Quote:

---

Originally Posted by Jonathan

Sorry I should have been clearer, I meant the accuracy of the straightness of the rail, not the tolerances they are machined to. Obviously I agree that overall profile rails are machined to much better tolerances and thus by far the most accurate rails. Round rail bearing blocks and supports are merely extruded.

---

Ok but even this is not quite correct.? They are very different in how they made and meant to be used.

Obviously unsupported are just straight bar with no support other than the ends so will flex in what ever direction the forces are applied.
Supported rails have the base support which dicate the strightness, again in my experience this base is far from straight or flat so if it bends or twist's so does the rail follow.

Profiled rail is meant to align to a registration egde on one rail. The rail is ground on at least 2 surfaces and in most case's all four. The rail actually as a special edge that's ground straight and true, often indicated with a arrow or machined grove/pattern. This edge is used to butt agianst the ref edge and is classed as the master rail while the other rails is left to float.
Now I know you know this but my point is that yes they are more far accurate just by the very nature of how they are meant to be used and all the ground square and true surfaces's. This is also what makes them very intolerent of uneven surface's and sloppy workmanship.

As soon as you properly fix down a profile rail it wont move, which is clearly an advantage, however if the rail is bent to start with you'll still have a bent rail.

I just checked the straightness of two profile rails and a SBR20 rail on my surface plate to check. Took readings 3 times for each over 400mm length in different places so should be reliable...

SBR20-400mm:0.03mm height variation.
Hiwin 15mm (760mm long): 0.06mm variation.
NSK 20mm (520mm long): 0.05mm variation.

The hiwin catalogue says +-0.1mm for normal precision height, but I don't think it states weather that is +-0.1mm for random rails, or +-0.1 over a single rail.

This agrees with my saying the the straightness is similar, but absolutely no reason to think SBR rails are better as they're clearly not!

Quote:

---

Originally Posted by Jonathan

SBR20-400mm:0.03mm height variation.
Hiwin 15mm (760mm long): 0.06mm variation.
NSK 20mm (520mm long): 0.05mm variation.

The hiwin catalogue says +-0.1mm for normal precision height, but I don't think it states weather that is +-0.1mm for random rails, or +-0.1 over a single rail.

---

So what you checking here Height or straightness.? Very different things.!! Also did you fasten them down along there full length.? Doubt you did and untill you do then the height check means diddly squat.!

Also your not 100% wrong about bent being bent just 50% wrong.???

Profiled rail won't tolerate any bend if it's to work correctly but that doesn't mean they won't come with a very slight bend to them over there length.? Hense the ground edge's so it can be registerd against a reference edge.
They are designed to flex ever so slightly so they can be manipulated perfectly straight and also why you have fixed master rail and a floating rail that gets adjusted by moving the gantry of what ever the bearings are attached too slowly across the rails length while tightening the rails fixing bolts, so tweaking out any bend or tight spots.! This is also why most profiled rails have 60mm centres between the holes.

Now if round rail is bent then it's bent and your stuck with it without resorting to off machine straightening. Often thou it's not the rail it's self but the rubbish base that's bent or twisted and often not flat on the mounting surface.

Profiled and round are very different beast's so can't and shouldn't be compared they are not in the same league.

Quote:

---

Originally Posted by Jonathan

SBR20-400mm:0.03mm height variation.
Hiwin 15mm (760mm long): 0.06mm variation.
NSK 20mm (520mm long): 0.05mm variation.

---

When comparing the parts yep but a more real world example would be that whilst installing on a machine you can further level the profile rails providing you've got the measurement tools. So those figures you've hinted at aren't set. They could even be made worse if you make a mess of the fixings which is less likely with the round rail as its only fixed both ends and not at regular intervals. Its as Jazz said, they're designed for precision and will give more of that over round rail but you have to exercise another level of precision when installing them to reap that.

I didn't fasten them down as the purpose of the check was to show that even before you mount the rails they are bent (I measured straightness for the profile rails) which is clearly true. The test was not intended as a realistic indicator of height accuracy in operation, as clearly either will conform to the mounting surface.

Sliding the gantry back and fourth and tightening the bolts to tweak out any tight spots also wont get the rails straight. It will merely copy any bend in the master rail to the subsidiary rail. Very few DIY builds rest the master rail against a reference edge, or indicate it to get it straight, hence for those any bend in the rail as shipped will remain (in one plane). For example if I just bolt down one of my 15mm rails to a piece of nice flat aluminum tooling plate, then use the sliding method to align and fix the other rail to the same plate, the error I measured will still be present so the measurement is relevant. That's what most people do and it's perfectly valid for supported rails.

Quote:

---

Originally Posted by Shinobiwan

Its as Jazz said, they're designed for precision and will give more of that over round rail but you have to exercise another level of precision when installing them to reap that.

---

Which is what I said originally; "the accuracy of a profile rail is not much different to a round rail - it depends how you mount them as how many of us correct the bend in a profile rail by comparing to an accurate straight edge reference?"

Quote:

---

Originally Posted by Jonathan

Sliding the gantry back and fourth and tightening the bolts to tweak out any tight spots also wont get the rails straight. It will merely copy any bend in the master rail to the subsidiary rail. Very few DIY builds rest the master rail against a reference edge, or indicate it to get it straight, hence for those any bend in the rail as shipped will remain (in one plane). For example if I just bolt down one of my 15mm rails to a piece of nice flat aluminum tooling plate, then use the sliding method to align and fix the other rail to the same plate, the error I measured will still be present so the measurement is relevant. That's what most people do and it's perfectly valid for supported rails.

---

In which case they have incorrectly installed them.!!!! . . . If the master rail reference edge is accuratly straight which it should be and is the whole point of having it then the subsidiary rail as you call it will be perfectly straight.!! . . If not used then you will have incorrectly installed them and defeated one of the main points of using profiled rails.!! Accurecy

I get the feeling Jonathan you haven't ever used profiled rails before.? Because if you had you wouldn't be so quick to argue with me about the bennifit of the master rail and having a straight reference edge. You would also know just how critical and intolerent the bearings are of miss alignment and poor mounting surface.!

If you have then you know I'm correct and pride is just making you deliberatly argumentative is all I can think.!!! . . . . either way your very wrong.!

I did not say they wont be accurate if mounted properly. I'm merely pointing out that most people don't mount them properly, and thus don't obtain the best possible accuracy. Nowhere have I said it's a good idea not to mount them properly.

We both already knew everything either of us has posted in this thread, so the only purpose in discussing is to let others know. Spreading the knowledge is the whole purpose of this forum, hence I discuss.

Your lasts post seems to be suggesting the reference edge of the master rail is perfectly straight? Which is surprising, as this simply is not true. The point of the reference plane on the rail is that is it parallel to the grooves the balls run in - you can't easily indicate off the grooves but can use a plane. You have to first straighten the master rail when mounting by attaching a bearing block and indicator and using a genuine straight edge as the reference for the indicator, or by clamping the reference plain to the straight surface.

Hiwin's document on the subject explains it all perfectly clearly, well worth a read:

http://www.hiwin.de/DownloadCenter/f...9TE14-1006.pdf

Quote:

---

Originally Posted by Jonathan

Your lasts post seems to be suggesting the reference edge of the master rail is perfectly straight? Which is surprising, as this simply is not true.

---

No you probably miss under stood.? I said "master rail reference edge" meaning the edge which the master rail butt's against be perfectly straight.

I've read the Hi-win document before and like wise fitted many hundreds feet of profiled rails, like tomorrow when I'll be drilling and tapping 160+ mounting holes for the 10mtr's of profiled rail so I feel I'm more than qualified to comment on all aspects of fitting them.!!

All that said and my original comment was that your statement that round rail and profiled are roughly the same regards straightness is incorrect. . . BUT for clarity's sake I'll ammend my statement to read "Rubbish total and utter rubbish" . . IF CORRECTLY FITTED.!!

Now lets go to bed it's 1 am. . :whistling:

Quote:

---

Originally Posted by JAZZCNC

I get the feeling Jonathan you haven't ever used profiled rails before.? Because if you had you wouldn't be so quick to argue with me about the bennifit of the master rail and having a straight reference edge.

---

I dont see that Jonathan has challenged the benefit of having a Master Rail or a reference edge directly, he has simply discussed the credentials of both and what those credentials will mean.

Quote:

---

Originally Posted by JAZZCNC

If you have then you know I'm correct and pride is just making you deliberatly argumentative is all I can think.!!! . . . . either way your very wrong.!

---

To be fair Jazz i think the shoe of pride fits the other man in the room, arguments drive discussions and that is what this is "a discussion on the fitness of unsupported rails for an application". So for me it is well done Jonathan for making full use of the forums.

You have both covered this topic very well and highlighted very good points, so thank you as i myself have taken away some knowledge from this discussion.

Keep up the good work guys as this is exactly what its all about - Education !

Quote:

---

Originally Posted by Lee Roberts

To be fair Jazz i think the shoe of pride fits the other man in the room, arguments drive discussions and that is what this is "a discussion on the fitness of unsupported rails for an application". So for me it is well done Jonathan for making full use of the forums.

---

Woooo who swallowed a dictionary then.!!!. . . . . Not me thats for sure. .:rofl:
Nah ME argumentative for the sake of pride. . .NEVER . . . Your wrong. :joker:. . . . . . I'm right thou has those who have ever fitted them "Correctly" will agree and wouldn't argue. :naughty:

Nite Lee.!

Okay thats better. Jazz and jonathan had been agreeing far too much recently.:heehee:

Bruce

Quote:

---

Originally Posted by JAZZCNC

Nah ME argumentative for the sake of pride. . .NEVER . . . Your wrong.

---

I wasn't referring to you Jazz, :apathy:.

Hi Jonathan,

Enjoying your stress analysis so far - good work there. I think people are finding the pretty colours more interesting than my old spreadsheet of numbers.

You mention the bed analysis - any plans for stress analysis on other areas of a cnc machine such as gantry sides?

Quote:

---

Originally Posted by routercnc

Enjoying your stress analysis so far - good work there. I think people are finding the pretty colours more interesting than my old spreadsheet of numbers.

---

Thanks - I used that spreadsheet a lot to start with.

Quote:

---

Originally Posted by routercnc

You mention the bed analysis - any plans for stress analysis on other areas of a cnc machine such as gantry sides?

---

Don't see why not. Although I'm not sure what to compare there? I could model different thicknesses of aluminium plate, but that wont tell us much more than your spreadsheet did. What would you suggest?

think i missed all the fun here-!

on my build i'm looking to mount the profiled rails to some 120 x 80 x 6 RHS, one side of the frame fixed, the other will be adjustable, and the frame will bolt together to allow for some tweaking and to shoehorn it into where i plan to install it.

will that give me enough adjustment do you think?

also Jonathan, what program do you use for your FEA simulations could be useful for my actual work

George

Hi Jonathan,

I was only thinking of various thickness plates vs RHS. Both cantilevered with one end grounded over an area representing a typical gantry end section (e.g 80x80 extrusion). For RHS the inside gap at the grounded end would also be filled in over the same area as the end of a typical gantry (e.g a spacer) since this represents the best bolted joint.

You are quite right that this is available in the original spreadsheet its just that your pictures are easier for most people to follow, and it is another question that gets asked on this forum. Just a thought since as this is now a sticky it might help newcomers to see the relative performance of common CNC machine designs if the analysis results are all grouped here.

Georgetheforge,
Sounds like a promising start based on the concept so far.

Allow me...

http://www.mycncuk.com/forums/showth...ign-principles

http://www.mycncuk.com/forums/showth...ail-Calculator

I had the links in my notes :)

Thanks ecat i was just about to go and find those links!

Just to let you know i have moved both those threads (and this one) to the FAQ's, Problems & Solutions forum as that seems the most appropriate place to have them at the momment.

Keep up the good work guys, without letting to much out of the bag, all of this kind of information is going to be grouped and made more readily available in the not too distant future. So thanks, as every little (or greater) helps towards what I am trying to do here for the site.

EDIT: Forgot to say i have also made them "Sticky" should you need to reference them at any time.

.Me

My little Roland uses unsupported rails, the Z axis rods are only fixed at the bottom end.

Perhaps it is a question of size, quite okay if you don't go wild on the length?

I got one of them somewhere ? Came from a school with dead controller, can jog but not run a program.

Quote:

---

Originally Posted by John S

I got one of them somewhere ? Came from a school with dead controller, can jog but not run a program.

---

I hacked it for USB serial and Windows 7. Means changing one socketed EPROM and cooking up a cable. You are welcome to try it :D