[4]

[How do I?] measure linear rail flatness

[3]
[10] Like what you see?
Click here to donate to this forum and upgrade your account!

cs900

maker of chips
H-M Supporter - Gold Member ($25)
Joined
Mar 21, 2016
Messages
223
Likes
222
#1
Ok, so a little background info. I'm in the process of converting my mill ( a PM45) over to linear rails, and i'm to the point where i'm getting ready to install the master rail for the y axis on the base. So I'm trying to measure the flatness of the master rail so i can then measure for parallelism on the slave rail.

So on to the good stuff...I think i've come up with a good method for checking flatness, but would like to hear you guys opinions. In my efforts i think i've come up with a poor mans repeatometer.


so as i slide the blocks down the rail the indicator will measure deviation from the block with the indicator to the other block.


so if I'm thinking this correctly, i'm getting a fairly consistent deviation of -.002 to -.003 along the rail which tells me my rail is bowed down at a fairly consistent rate. So thoughts, is there a better way to do this? Am i completely off my rocker?

guess i should also add, that my surface plate isn't large enough to fit the base...and also the base is about 180lbs so moving it is a bit of a chore....
 

Bob Korves

H-M Supporter - Sustaining Member
H-M Platinum Supporter ($50)
Joined
Jul 2, 2014
Messages
5,655
Likes
5,966
#2
Gauges like that one (and the Rahn Repeat-O-Meter) do not show flatness or level. They simply show changes from one area to another (whether the path is moving more up, more down, or staying the same), which must then be interpreted. They do not know level or flat. To get a bit of understanding, imagine the rails were warped into a perfect vertical circular hoop, 50 feet in diameter. The gage would move along with a steady needle all the way around, but the surface would not be flat, correct? Gages like that one need to be used in conjunction with another type of measuring tool. To use your surface plate example, it would typically be an autocollimator, which compares angles very accurately. Together, and with charting all of the results, the readings suggest the actual 3D topography.

Still, very nice job of rolling your own test rig! Now you need the poor man's autocollimator...
 

cs900

maker of chips
H-M Supporter - Gold Member ($25)
Joined
Mar 21, 2016
Messages
223
Likes
222
#3
Thanks for the reply bob. Yeah, i tried to word the errror i was seeing to reflect that i measured a convex bow in the rail since each "step" i took away from the back side showed a negative value, that just happened to be fairly consistent along the length.

So after thinking about it for a while i think i came up with a better approach, but absolutly would like to hear feedback still. My reasoning was that if the base is indeed bowed i could loosen all but the back bolt which would allow the rail to become straight again. So i did, and using the same method as before verified no real deviation from point to point on the rail. I then went back and measured the deviation using the rail as my straight reference. 20180119_004810.jpg
Sweeping to the end net me about .004 deviation, which seems more realistic than what i was seeing before.
20180119_004820.jpg
So this confirms the convex bow in the rail i was seeing before, and i think i can just shim the rail based on the deviation measured from this method. Thoughts?

And im absolutly going to look up the autocollimator and see what i can rig up, haha.
 

4ssss

Active Member
Registered
Joined
Apr 25, 2017
Messages
451
Likes
262
#4
Can I ask where you got those rails? I have a job going on now that needs a heavier set than what was given to me.
 

cs900

maker of chips
H-M Supporter - Gold Member ($25)
Joined
Mar 21, 2016
Messages
223
Likes
222
#5
the rails are NSK LH20. I actually got these used off ebay, and was super fortunate as they look like they were never used. They also have these cool lubricating strips sandwiched under the endplates on the blocks (the little tan pieces) that according to NKS are all the lubrication the blocks need.
For what it's worth I got them from seller athomemarket
 
D

Deleted member 473

Guest - Please Register!
Guest - Please Register!
#6
One thing you didn't do when you laid the rail down on the bed. You need to take your feeler gages and check for gaps under the rail and the bed. Your bed is more than likely not flat from wear in areas causing the readings you are seeing. Linear rails are extremely flat and straight right out of the package. You will probably have to shim up under the rails with different thicknesses if shim stock from 0.0005" to 0.0040" in thickness. If you know how to scrape and fit slides, you can scrape the bed to a flatness using the rail as a straight edge.
 

cs900

maker of chips
H-M Supporter - Gold Member ($25)
Joined
Mar 21, 2016
Messages
223
Likes
222
#7
One thing you didn't do when you laid the rail down on the bed. You need to take your feeler gages and check for gaps under the rail and the bed. Your bed is more than likely not flat from wear in areas causing the readings you are seeing. Linear rails are extremely flat and straight right out of the package. You will probably have to shim up under the rails with different thicknesses if shim stock from 0.0005" to 0.0040" in thickness. If you know how to scrape and fit slides, you can scrape the bed to a flatness using the rail as a straight edge.
that's a good idea. I'll try checking it with feeler gauges tonight. I tried shimming it this morning based on the error I saw, but doing the ghetto repeatometer test again showed the same bow in the rail after shimming.

I have the blueing compound and a carbide scraper from when I fitted the gibs to the mill maybe I'll give that a go.
 
D

Deleted member 473

Guest - Please Register!
Guest - Please Register!
#8
Another thing to do along with what I said above.
Start in the middle of the length of the rail. With all of the cap screws loosen, Place your dial indicator next to the cap screw, on top of the rail. Zero the dial with the cap screw loosen. Next tighten the cap screw and note the amount of movement. You should get somewhere around.004" as you indicated above. Place a .004" shim under the capscrew between the bed and bottom side of the rail. Repeat tightening the cap screw. You should get 0.000" reading. Get it within .001" and move down the the next cap screw and check.

When tightening cap screws holding a rail down. Start in the middle of the length and work outwards to the ends, just like you would do installing a head on a engine. This should get you the best of accuracy and alignment of the rails when done.

Ken
 
D

Deleted member 473

Guest - Please Register!
Guest - Please Register!
#9
If you can get access to a surface plate, do so, and check for flatness of the backside of the rail. Easy to do, place your indicator at the middle of the length of the rail. Press down on the rail and see if the pointer on the indicator moves. Let's hope it don't. If it does, you have to take the readings in account to fitting to the bed. Don't try to straighten a rail! Your best to buy another one to take its place.

Don't rely on the bed of your lathe to be flat!!!! Just looking at the pictures, part of it is machined, the old ways could be scrape to a known flatness? I would not trust it without verifying with a know flat straight edge. And even then, does it have a twist in the bed? A lot to think about here. Take it a step at a time. You're doing something I've wanted to do is build a CNC lathe using linear ways. And using a existing bed is a good start as long as you know it is straight and flat without a twist.

Let us know how you make out. We're here to help.

Ken
 
Last edited by a moderator:

cs900

maker of chips
H-M Supporter - Gold Member ($25)
Joined
Mar 21, 2016
Messages
223
Likes
222
#10
interesting. The NSK installation manual indicated you should start at one end and tighten the bolts in a row.

I do have a surface plate, and I'm going to check the rail on my lunch break.
 

cs900

maker of chips
H-M Supporter - Gold Member ($25)
Joined
Mar 21, 2016
Messages
223
Likes
222
#11
ok...so one long lunch later I'm feeling better about things. I took the rail onto the surface plate and the total deviation was under .002" cool...it's as flat as I could ask for.

So I put the rail back on the base and tried Ken's method. Indicator right next to the bolt head, and starting in the middle of the rail and worked outward.


All but one of the heads was within a few tenths, and the other one was about .001 low. I kind of knew it had a low spot there as when i stoned the top the lubricant would always pool there. Cool...so after adding the appropriate shims I was curious and went back to the repeatometer trick. The whole rail is within .001 now. Cool! Starting to feel confident that it's actually flat.

so now onto measuring parallelism to the slave rail!


I swept the surface and found the high spot, set my zero and went to work. Not the best results I'm afraid. It's low as much as .010 in one area! But I feel a little lucky, I stuck a straight edge across the top of the master rail and measured the change in height to the high spot on the slave and it's within .005". Well within NSKs advertised limit. So I guess I have some shimming to do, and then onto the X axis
 
D

Deleted member 473

Guest - Please Register!
Guest - Please Register!
#12
Before you jump over and shim up the slave rail by .010", measure the height of the surface to the center portion of the slide you have there and see if the mounting surfaces are at the same height. If they are, that means in my book that you are ok and don't shim the rail. The NSK bearings will have a slight tilt side to side. Not easily notice at the rail, but when reach out over the slide base as you show in the picture, it's multiplied by the length of the arm of you dial indicator mount. Wild guess, if that's 10" reach there, then 0.001" movement at the NSK bearing, that's .010" drop at the dial indicator. Think about it a bit. Ken
 

cs900

maker of chips
H-M Supporter - Gold Member ($25)
Joined
Mar 21, 2016
Messages
223
Likes
222
#13
Ken, maybe I'm not following your logic.

I've tried to measure the height difference from the top of the master rail to the top of the slave rail the best I could. this was around .005 difference. I did this at the highest spot on the slave rail in order to be able to shim the rest of the rail up. So once I did that I zero'd the indicator at the high spot and swept the rest of the rail to get the deviation from that point. My thinking is if the master is straight, and I pick a point on the slave, any deviation from that point is the change in height from the master. Or am I overlooking something?

I also did check to see if the bearing was tilted a bit. To measure the height from the master to the slave I placed an Al extruded bar (I know I know, it's the best thing I had long enough) on top of the master block and measured from the top of the bar to the top of the master rail near the bearing, and then again to the slave rail. I then took off the bearing, flipped it around 180* and measured again. Only a few thou difference. This tell me the bearing is pretty darn flat, but I suppose I could still have a twist in the rail.
 

cs900

maker of chips
H-M Supporter - Gold Member ($25)
Joined
Mar 21, 2016
Messages
223
Likes
222
#14
but I also see your point. If my master rail still reads .001 off in spots it will certainly exaggerate the reading at the slave side.
 

Bob Korves

H-M Supporter - Sustaining Member
H-M Platinum Supporter ($50)
Joined
Jul 2, 2014
Messages
5,655
Likes
5,966
#15
If you look at the photo of the rails on the ways, with the indicator and stand over the top, do you see the flat areas between the dovetails with the numbers written on it? That is the original, unworn, untouched factory reference surface. They made that nice machined area for a reason. Lightly stone any burrs off the surface, and then use a depth mic to measure from there down to the ways. You are over thinking this. Direct measurements are always preferred. The introduction of intermediate stuff just clouds the issues and offers more places for errors to accumulate.
 

magicniner

H-M Supporter - Silver Member
H-M Supporter - Silver Member ($10)
Joined
Oct 21, 2017
Messages
431
Likes
382
#16
You might be measuring Twist in your primary when you see vertical deviation in your secondary.
 

Wreck™Wreck

Silver
Former Member
Joined
Sep 29, 2014
Messages
2,102
Likes
1,732
#17
This is when you actually need a surface plate.
 

cs900

maker of chips
H-M Supporter - Gold Member ($25)
Joined
Mar 21, 2016
Messages
223
Likes
222
#18
If you look at the photo of the rails on the ways, with the indicator and stand over the top, do you see the flat areas between the dovetails with the numbers written on it? That is the original, unworn, untouched factory reference surface. They made that nice machined area for a reason. Lightly stone any burrs off the surface, and then use a depth mic to measure from there down to the ways. You are over thinking this. Direct measurements are always preferred. The introduction of intermediate stuff just clouds the issues and offers more places for errors to accumulate.
Yeah, you're right. This is what I ended up doing and saw only .0005 difference from left to right side. Although interestingly enough there was a slope from front to back. Not a problem as I can shim the column to get square.

I wish i had a surface plate large enough to fit the whole base, but unfortunately I do not.

So looks like that issue is solved and I'm back onto the retrofit!
 
R

Robert LaLonde

Guest - Please Register!
Guest - Please Register!
#19
Ok, so a little background info. I'm in the process of converting my mill ( a PM45) over to linear rails, and i'm to the point where i'm getting ready to install the master rail for the y axis on the base. So I'm trying to measure the flatness of the master rail so i can then measure for parallelism on the slave rail.

So on to the good stuff...I think i've come up with a good method for checking flatness, but would like to hear you guys opinions. In my efforts i think i've come up with a poor mans repeatometer.


so as i slide the blocks down the rail the indicator will measure deviation from the block with the indicator to the other block.


so if I'm thinking this correctly, i'm getting a fairly consistent deviation of -.002 to -.003 along the rail which tells me my rail is bowed down at a fairly consistent rate. So thoughts, is there a better way to do this? Am i completely off my rocker?

guess i should also add, that my surface plate isn't large enough to fit the base...and also the base is about 180lbs so moving it is a bit of a chore....
Large precision granite square or straight edge and thin shim stock or feeler gages will get you close. Then a flashlight to finish. No kidding. A 24" of mediocre quality granite straight edge is only a few hundred dollars. I have a small granite square I use for checking perpendicularity, but have not yet picked up a longer straight edge for checking flat.

A large granite surface plate might work, but you would have to 4 corner level the base first before trying to measure flatness. I assure you that cast base is not likely to be perfectly perpendicular between its precision top surfaces and its feet.
 

cs900

maker of chips
H-M Supporter - Gold Member ($25)
Joined
Mar 21, 2016
Messages
223
Likes
222
#20
yeah, I was a little hesitant to use the top surface just for that reason. But in the end, i'ts the best I can do without a surface plate. I also filled the base with epoxy granite, so I'm hoping that will help resist any deformation in the base when I bolted it down.
 

Wreck™Wreck

Silver
Former Member
Joined
Sep 29, 2014
Messages
2,102
Likes
1,732
#21
A large granite surface plate might work, but you would have to 4 corner level the base first before trying to measure flatness. I assure you that cast base is not likely to be perfectly perpendicular between its precision top surfaces and its feet.
Indeed, we have a 4' X 8' X 24" granite plate, excellent for such work. Probably not accurate enough for hobby work however.

i-mWR2FT3.jpg
 

cs900

maker of chips
H-M Supporter - Gold Member ($25)
Joined
Mar 21, 2016
Messages
223
Likes
222
#22
wow, that thing is thick!
 

KBeitz

H-M Supporter - Gold Member
H-M Supporter - Gold Member ($25)
Joined
Apr 17, 2018
Messages
385
Likes
192
#23
Shoot or skim a laser across the top...
 
[6]
[5] [7]
Top