View Full Version : Clever ways to maintain parallelism?

08-25-2009, 04:23 PM
Suppose I have a flat plate of aluminum and I'm going to bolt two rectangular blocks to the plate. I need to keep the faces of the blocks parallel to each other. The distance between the blocks is 20" and the blocks are 0.25" X 0.75" X 1.75"

I don't have a 20" mic ... :) I've thought of a couple of ways to do this, but they all involve a good bit of time and setup. I thought someone here might have a clever solution for me.

08-25-2009, 04:26 PM
clamp a straight edge to the adjacent sides???

08-25-2009, 04:27 PM
20" +/-??? parallel within +/- ???? tolerance is critical number. Determines the method to use.

The Artful Bodger
08-25-2009, 04:33 PM
Tip the plate vertical and use levels on the blocks?

08-25-2009, 04:36 PM
Oops - thought I mentioned that...

They need to be parallel within .0005" - these blocks will form resolving slits in a parallel plate energy analyzer for 5 KeV electrons.

08-25-2009, 04:49 PM
Clamp the blocks together and then drill, tap and fasten them on one end to the centre of a reference piece of material with a good edge at least 8" long. Saw reference in half to separate blocks.

Fix them on the plate in position and use a straight edge to align them via the edges of the piece(s) you fastened to them. Remove reference pieces.

[edit] I omitted a step. After clamping the blocks together machine the surface that will be fastened to the reference part. It doesn't make any difference if that surface is square to the others as long as it is flat.

Btw, wouldn't the Mossbauer effect be applicable here?

08-25-2009, 04:58 PM
if the surface plate is big enough...bolt the blocks in place, more or less, with a couple of cap head screws.... clearance between screw and holes is your 'adjustment'. set it up on the surface plate, on it side. Use a precision square (a good one!) to get each face square to the plate. tighten the screws, check, drill & ream for dowel pins. if the plate is to small, scrape/file the edge of the plate so its flat & straight (not ideal, but can be done with a plate less than its long) then set up the block using the precision square against the edge.

this assumes the .0005 is on squareness, not on the 20" dimension...in which case you have to come and borrow my 20" mic :)

08-25-2009, 05:01 PM
You have a lathe (or access to one there at school). As such, I would take a piece of straight bar stock maybe an inch in diameter and face one end. Part it off close to 20" and hold the work between centers and use a DRO (got one of those?) to start to face the other end to exactly 20". Pull it back in the chuck and face it off using the start you cut as the reference. The two ends should be as parallel as the lathe is capable of making them. Clamp around a block, your 20" parallel stick and another block using a bar clamp and set the works on the plate. Slightly loosening the big bar clamp will allow the blocks to drop to the surface of the plate, and snug the clamp back up. Leave your test bar in place while doing the rest of the mounting work and then remove it.

It works for me every time I need to precisely place two blocks 20" apart and parallel on a plate:D


08-25-2009, 05:50 PM
Even though the OP didn't say if he had one, I was wondering if the ways on a knee mill are accurate enough to use table travel and a DTI to check for parallelism of the block faces? This is assuming the ways are in good condition. You might even be able to get the 20"dimension fairly close this way also with an edge finder but given the inaccuracies in the lead screw from wear and manufacturers tolerances, it might not be as accurate as you require. Just wondering....

08-25-2009, 06:00 PM
Align and fasten the plate to the table of your best vertical mill.
Locate and bolt the first block in place, aligning it with the cross feed.
Leaving the indicator set move the table 20 inches observing all the rules of backlash. If you have a DRO the job is easier.
Indicate and bolt the second block.
Blocks are parallel and 20 inches apart.
Move the entire assembly to a different spot on the table and retest. This will give you some measure af assurance that the method is reasonably sound. If you find wild differences then the mill is in really bad condition and you should look for a different solution.


08-25-2009, 06:03 PM
Some guys at work do this stuff all the time to a few tenths, except it's 1.5 meters of separation.

What they do is mount a DTI on a long tube. One end of the bar has a ball hot glued into it. By holding ball end of the bar against one of the blocks to be aligned, and then sweeping the DTI end, you can take a point measurement of the separation. Then you move the bar, and sweep again. Presto - the difference between the two measurements tells you how much you need to move it. You can put three point/ball feet on the bar to help repeatability. Again, hot glue and some holes drilled in the side for cheap bearings is all you need.

It's pretty quick, considering you only need two points to do a pretty decent alignment (one at each end of the surfaces to be parallel). Then, you just do sweeps in between to verify it after you're fairly confident they're straight.

08-25-2009, 07:48 PM
Take two pieces of 1/2 in sq rod, clamp them together side by side, mount them square to the mill table, then mill across one end. Reverse the package without disturbing the clamps holding the bars together, then mill across the other end. Let the temperature stabilize at room temp, then trim to the required length. Separate the pieces, debur, and you have two equal length spacers.

Use these between your blocks, laying out the four parts as a perfect rectangle so each bar keeps the blocks spaced to its exact length.

Unless I'm missing something, your biggest problem is getting the spacers to whatever exact length you need. You may not have a tool capable of measuring accurately at that length. If your need is mostly to have the faces of the blocks parallel, then it's down to how well you set up the blocks with the spacers, and then keeping that alignment as you tighten the fasteners.

Uncle O
08-25-2009, 08:05 PM
Dowel Pins

doctor demo
08-25-2009, 08:27 PM
Two Twenty inch stacks of jo blocks.


Forrest Addy
08-25-2009, 08:31 PM
Boy! There's so many sources of error possible I can hardly list them. Aluminum is handy and machine's nice but the plate. How thick is it? It is flat/ How is it held. Is gravity a source of potention distortion/deflection? What are the three dinemsional considerations? Does the part heat in service?

A sketch or some form of visual laid woult sure help.

If the system is rigid and you have a good granite flat there's no reason that with care and controlling the environmental variables you can't hold tolerence limited only by the plate's accuracy. That is if you get cast iron control of the variables.

08-25-2009, 08:48 PM
Interesting ideas!

My first thought was to use a milling machine, but I don't have access to a mill with 20" of travel. The Bridgeports around here are 18 by 36? or something like that (i.e. ~9" and ~18" of travel).

I also thought about making a bar and clamping my DI to one end. It seemed like it would be a finicky job and not be very repeatable, but hearing that other people have had success with it is encouraging. I like the idea of using three ball bearings.

A 20" test bar would be nice too, but it requires that the bar be exactly the right length. Since I don't have a mill that is large enough to do it one shot, I was planning on drilling and tapping the plate for the blocks. Then I could adjust the blocks slightly to gain parallelism. Knowing me, I would make about 20 test bars, each one just a tiny bit too short ... :D

Using a straight edge is an interesting idea, I had not thought about that.

The exact distance between the blocks is not critical, just the parallelism.

Evan - these are 5 KeV electrons so they do not interact with the nuclei of atoms, i.e. they do not cause gamma emission.

Forrest - the plate is pretty wimpy. It is 0.25" by 8" by 24". This will be held at ultra high vacuum, so the materials were limited to SS and aluminum. The SS proved to be cost prohibitive. It will be part of a larger assembly and deflection should not be an issue (there is no load and it is well supported at a 45* angle). Heating is not a major issue, either.

08-25-2009, 10:40 PM
Stick a first surface mirror to each block with some vacuum grease and adjust for maximum "hall of mirrors" effect. It's extremely sensitive.

08-26-2009, 12:40 AM
Evans 'hall of mirrors' approach will work extremely well. I don't know what the absolute error possibility is but it would be easy enough to check experimentally. Just mount two blocks on a mill table close enough to measure with available instruments. Adjust as Evan said for maximum hall effect and then check the error.

Sounds like it might be a simple solution.


08-26-2009, 03:18 AM
i'd use cast plate, the rolled stuff does wierd distortion with temp [as i found!] cost more but stable, i liked Evans 'hall of mirrors' answer, i suppose its an optical lever for want of a better description, 'light is tight, as a brum engineering instructor i was subjected to used to say [with reference to accuricy [usual retort of scotts is tighter]

08-26-2009, 05:25 AM
Fix one of the bars to the plate, abut the other bar alongside fix a mirror at 45 degrees at the end of each bar so they are facing each other on the top bar(loose) and apposing on the fixed so that a beam will travel to the first mirror reflect to the non fixed bar then travel along the bar to the other end back to the fixed bar adjust the mirrors so the beam is visible on a target. Now when you move the other bar when the beam is in the same position on the target the bars are parallel.
All a bit fiddly but should work well you jsut have to get the 20" right.


08-26-2009, 09:03 AM
There are a number of optical approaches that will work well. By using a couple of beam splitters and a couple of diffuser lenses the parallelism can be nominally set to within a wavelength of light via fringing which is about 0.00004". Beam splitters can be made using pairs of binocular prisms stuck together with oil which will give about a 50% reflection at the interface. Diffuser lenses can be had from camera viewfinders of disposable cameras.

08-26-2009, 11:14 AM
The optical approach might sound esoterically difficult, but Tom is in the Physics department and might just have access to lots of toys that would make this easy.

As for the test bar....you said you would likely keep making them too short and then said that distance is not as critical as parallel. I was advocating for making the bar on the lathe....and making two ends parallel should be really easy with that tool...provided the bar does not warp or something. by definition the faced end of a bar will be perpendicular to the axis of rotation. Even if the lathe is a bit off and makes the end cupped, it should still meet your needs.

Forrest though points out an important issue...when you need parallel to .0005" across 20" of space, you have potential for lots of change as the plate warms and cools etc. This is sort of esoteric stuff but potentially very important depending on the function of your fixture as what you have now may not be what you have in use. If the plate is not perfectly flat to start with, you will never be able to get the works square and parallel either. No fixture can undo what happens when you bolt the two blocks to a plate with a curve across 20" of space. Making up for surface imperfection like that would mean you have to scrape the bottom of your blocks until you get perpendicularity of the blocks in spite of the curvature of the plate.

Edit-- with this in mind, I would think that attaching the blocks and machining them parallel in place would make the most sense. The issue is the work envelope of the machine you would need to do this

Paul Alciatore
08-26-2009, 11:43 AM
Some guys at work do this stuff all the time to a few tenths, except it's 1.5 meters of separation.

What they do is mount a DTI on a long tube. One end of the bar has a ball hot glued into it. By holding ball end of the bar against one of the blocks to be aligned, and then sweeping the DTI end, you can take a point measurement of the separation. Then you move the bar, and sweep again. Presto - the difference between the two measurements tells you how much you need to move it. You can put three point/ball feet on the bar to help repeatability. Again, hot glue and some holes drilled in the side for cheap bearings is all you need.

It's pretty quick, considering you only need two points to do a pretty decent alignment (one at each end of the surfaces to be parallel). Then, you just do sweeps in between to verify it after you're fairly confident they're straight.

This sounds real good to me. You could test all along the length with such a device.

08-26-2009, 11:49 AM
I'm wondering if you might not be able to do it by drilling and reaming each block with 2 dowel pin holes with a few tenths of negative clearance and then drilling the same dowel pin hole patterns in the plate using a single machine setup and as few table moves as possible. It seems like you should be able in principle to get as parallel as 2 y axis moves of the machine can be.

I've never tried anything with this accuracy requirement but I have to say I like dowel pins generally for the molds I've made for my research using theory from an old diemaking book!


08-26-2009, 12:00 PM
Why not put a .0005" or better dial inicator on a rod and use the rod and indicator as a comparator at the four corners of the two blocks? Basically it's an inside mike that reads with a dial indicator rather than a micrometer head.

08-26-2009, 01:18 PM
Paul, the test bar would work very well but I would be mounting the blocks on the plate before hand with only a small amount of play. So the 20" isn't important, but what is important is matching the actual distance between the blocks to the actual distance of the bar. I may end up going this route and cut some slots to allow the blocks to slide just a little bit. (as it is, they can "twist" about .005" from one corner to the other) I could then get them all set up, tighten down the screws and then drill for dowel pins.

These blocks actually form the back edge of some resolving slits. Even if I had access to a mill with the required amount of travel, machining it would be very difficult since the block tapers down to a knife edge for about .5" along it's length. The idea is to cut this taper and then skim a very small amount of the top of the entire block, ensuring that the edge (now no longer a knife edge) is flat with the remaining material. Then I will reference off of the remaining material. This evening I will post a picture, since this description is probably not very good. Perpindicularity to the surface is unimportant. I just need the two slits to be parallel.

Cameron, using the mill table would work well also, but I don't have access to a mill with 20" of travel. :(

Mark - I had not considered cast plate. In the past, these devices have always been made from SS or 6000 series aluminum due to vacuum considerations. I'm not sure how cast aluminum behaves in vacuum or how conductive it is. I will have to do some research and see what I can learn.

Thanks for all the suggestions! I had a lot of different ideas bouncing around in my head but, seeing as how I'd never tried it before, I didn't know which one would be best. It seems like quite a few of them would be equally good.

The optical approachs are interesting - it seems like it might take a little longer to set up but it might be fun to try. That was something I had not considered.

08-26-2009, 03:34 PM
My suggestion was not to use the bar as a measuring device but as a positioning fixture. If you can get a straight bar maybe an inch in diameter or so and face each end, then each face is perpendicular to the length of the bar. If you place a block at each end of the bar and clamp around the whole works, then you have two blocks, 20" from one another and parallel. Carry that whole works over to your plate and set it down. The two blocks are now 20" from one another and parallel. Attach as needed and then remove the clamp and bar.


08-27-2009, 10:49 AM

With the dowel pins, assuming that holes are the right size etc, I don't think you need need 20 inches of table travel. If you can rotate the column pivot without hosing the tram, you can drill 2 sets of holes that are parallel by moving the head without moving the table much or at all in the x direction as long as you don't need perfect location in the y direction.

Drill holes spaced so many inches in the y direction for the blocks exactly parallel to the edges of interest.

Mount the plate on the table surface with the head pivoted to the left. Drill 2 holes in the y direction with the same spacing as is in your blocks. Pivot the head to the right. Drill 2 holes with the same spacing in the y direction. The exact y coordinates of the ends of the blocks will not be aligned but the relative y coordinates between the holes will be correct. Thus you'll get 2 sets of holes that are exactly parallel but may be slightly offset. You can pin the first block and use a straight edge to get an approximate y position to drill the holes for the second block.

If a slight y offset is acceptable then you should have two parallel blocks as long as the holes originally drilled in the blocks are perpendicular to the edges you wish to make parallel. This assumes that the y axis of your mill stays where it is supposed to and that the dieties involved in reaming holes to get the right negative clearance for dowel pins are smiling on you.

This is a thought experiment result based on my experiences with dowel pinning smaller parts. It may not even be a good way. Your mileage may vary.


08-27-2009, 03:53 PM
If you really don't care about the length being exactly 20" (+- .060) then how about taking a piece of stiff tubing (1/2" to 3/4" diameter or so) and weld/glue/braze/etc. a ball bearing on each end.

I like the idea too of just mounting a dial indicator on the end of a rod.

08-27-2009, 08:39 PM
Dave5605 has one of the simpler, better ideas as you end up with a rod that is similar to a micrometer standard with spherical ends.

For 0.0005", your plate need to stay flat too. Any warp or bow whatsoever and you kiss the 0.0005" goodbye.

If you have a Starrett or other good quality 24" machinist's rule, using the 20" graduations and a couple of ground blocks (gage, 123, etc.) as followers, eyed with a proper magnifying lens, could get you close to the dividing accuracy of the rule if the 20" span needs to be close.

Ooh, clamp the ground blocks to the rule, using the magnifier to spot the graduations accurately. Now you have a giant GO gage. :D