machinist@lists.ibiblio.org
Subject: Machinist
List archive
[machinist] Making Accurate Straight-Edges from Scratch by John A. Swensen - Page 2
- From: Lawrence London <lfljvenaura@gmail.com>
- To: machinist@lists.ibiblio.org
- Subject: [machinist] Making Accurate Straight-Edges from Scratch by John A. Swensen - Page 2
- Date: Fri, 2 Jan 2015 00:04:00 -0500
http://www.practicalmachinist.com/vb/general/making-accurate-straight-edges-scratch-john-swensen-295849/index2.html
+ Reply to Thread
Thread: Making Accurate Straight-Edges from Scratch by John A. Swensen
-
Web evangelists and archivists should salivate at the prospect of building such a knowledgebase. With today's software technology it could be done in style, with ease using open source apps and programming tools. The PM database could be marketed to some company (an exchange agreement, mutually beneficial), possibly a Public Benefit Corporation (PBC), for conversion to such an online resource by skillful and imaginative developers.
Originally Posted by Richard King
It is so interesting when you share the stories of where you worked. Thanks. It would be a great service to all or to the history books if all of the machinists were to share all these stories so younger generations can read about them.
Don deserves considerable credit for starting and maintaining this quality forum which is without equal anywhere.Have to Thank Milacron for giving us this forum! RichRichard King likes this. -
I can't see myself needing a straight edge longer than 3' Rich, and the chances are if I was doing work longer than that I would be using either the autocollimator or (less likely) the alignment telescope. It was more just musings in that I can't say I've ever seen a straight edge such as I described but can't see why it shouldn't be successful. I too have seen numerous precision machines that compose either entirely or largely of weldments, and the reasons for their lack of more widespread use was unrelated to the thermal expansion of the steel used, and more to do with very different advantages of CI in these situations. Likewise I've seen steel, in various sections, used as straight edges, but the thought of scraping one, especially a long one, doesn't exactly thrill me. The ones I've seen have also been commercial sections of available steel instead of specifically fabricated weldments to take advantage of the appropriate mechanics and engineering in straight edge design.
Just considering a couple of points that have been raised, I'd agree, a weldment typically contains significant stresses as a result of the welding, and I looked into the possibility of stress relieving the weldment. It would still be significantly cheaper than a casting, even if the stress relieving was done commercially. However on further consideration I wondered if the stresses are in fact such a disadvantage? These stresses are latent, that is to say they are of no significance unless opposing stresses are relieved. In fact a stresses structure can be made to be significantly stiffer and stronger than an unstressed structure, and it's one of the most fundamental tenets of civil engineering.
With regard different expansion rates of different materials, I think some are allowing themselves to become lost in the science and not considering how this will be used in reality. There is little to no temperature differential to be considered here, other than that introduced by the user through misuse. You don't take your fabricated straight edge from on top of the heater and drop it on the surface plate and start going to town on it! It may mean such a fabrication may take extra vigilance to ensure that undue heat isn't introduced in to it, but I would not expect this to be a factor. Furthermore, what I had in mind was largely a steel fabrication, with only a relatively thin veneer of cast iron affixed to the base to make it easier to scrape. I would think by using a conventional camel back design, and having the arch stressed, a relatively long straight edge could be fabricated that would be considerably lighter and more manageable to handle, while being significantly stiffer than the equivalent cast iron alternative.
Finally, I considered steel as the solution to the stiffening structure (as that's all this is, and it's the cast iron "shoes" that are the business end of this), as that's what most here would be most comfortable with, including myself. But with really large straight edges, weight becomes a significant concern. Which led me to wonder if even steel needs to be used? Could a carbon composite "camel back" for example be designed that would lead to quite long and (relatively) light weight straight edges? That's definitely something that I don't have the ability to easily fabricate, but it certainly does spark the imagination as to the potential to think outside the conventional box that all straight edges must be wholly cast iron construction lest the world stop turning as a result of using something else. It seems in this area there's been little to no progress in terms of alternatives since the days when cast iron was used for pretty much everything, and for manufacturers with a casting background little or no incentive to do so. However in my experience the ready ability to have high quality cast iron castings done isn't as available as it may have been, say 50 years ago, certainly not around here anyway. The cost to take a concept of a straight edge drawing through the design, pattern making, and casting stages is also very significant, with many foundries disinterested in one-offs also. Maybe it's time to start seriously exploring alternatives?Richard King likes this. -
Yes, if you are comparing three straight edges, twist can be a problem. The condition for generating three flats from scratch is often misquoted and in error. Properly stated it goes like this, The only surface that can be generated when three areas are scraped to intimate contact when compared in all combinations AND ALL ORIENTATIONS is a plane. Originally Posted by Mark Rand
He doesn't seem to show how to avoid twist...
It is not sufficient to just compare them in all three combinations (1:2, 2:3, and 3:1). The reference above shows this in a one dimensional manner and it seems to work. But all flats are not one dimensional. They are two dimensional and that allows for twists which would not produce true flat surfaces.
Straight edges are relatively narrow and would only allow 180 degree or end-to-end rotations. This also does not allow you to reliability remove twists. This is illustrated here:
As you can see in the top, three surfaces with the same twist can come in intimate contact at all points. And in the second section you can see that rotating one by 180 degrees (end-for-end) still allows that same, intimate contact, even with a severe twist.
In the bottom illustration you can see that by introducing a 90 degree rotation you can detect these twists and eliminate them with proper scraping or other processes. So the three surfaces MUST be tested both in all combinations and in, at lease two, perpendicular orientations. I would start testing with rotations very early in the process to ensure a good start.
These perpendicular orientations are not easy with the narrow straight edges so it is not a good idea to try to use three straight edges to produce flat ones. It is better to use three square or round plates and make them flat with proper testing. Then use one or more of them as a reference to scrap your straight edge. The plate would not be scraped any further in the process of scraping the straight edge.
Now, if you have three straight edges that have already been scraped flat and you are just doing touch-up work, it MAY be possible, with a lot of care, to do this touch-up without a reference flat. But you have no real assurance that this is the case. In other words, if it does come out OK, it is just dumb luck. And if this were done multiple times with the same three straight edges, then the probable error would increase with each such procedure.
Do not get the idea that only two surfaces are necessary from the third section of my illustration. This illustration was designed to show the effects of twist only. With only two surfaces you can indeed eliminate all twist, but they may have opposite and equal spherical shapes: one convex and the other concave. This is, in fact, how optical lenses and mirrors are generated: only two pieces of glass are ground/lapped together, with rotations to prevent twists and a spherical surface is generated. Usually the bottom, stationary surface becomes convex and the top, moving surface becomes concave. The depth of the curve is measured at intervals and the rough lapping is stopped when it is close to the desired radius. If it becomes too deep, the two pieces are swapped, top and bottom and then the curve decreases. Generally you move around the stand that the lower surface is mounted on after a few strokes and also rotate the upper surface in your hands at the same time. These rotations are kept somewhat random so that no pattern can be established. Complete regularity is not a virtue here. The law of averages rules. I have made a telescope mirror with this process and it works well. When this process is mechanized for large telescope mirrors, the hard part is to make a machine that has enough slop in the motions to eliminate any effects produced by too regular of a motion. The most precise surfaces are made with machines that are deliberately made sloppy.
But to generate a flat, THREE surfaces are needed. -
01-01-2015, 07:50 AM #24
- Join Date
- Mar 2014
- Location
- St Paul, Peoples Republic of Minnesota
- Posts
- 42
The three plate method was invented by Joseph Whitworth when he was an apprentice under Henry Maudsley in London in 1770, Maudsley invented the screw cutting lathe, Whitworth became the owner of a huge manufacturing machine shop, he made breech loading cannons in calibres up to 12 inches
For the British Navy. Both of these men were mechanical genius's in their time. -
Surely any twist would become apparent just by taking a print with the top straight edge rotated just enough so the left corner of the top straight edge was over the right corner of the bottom one?
-
01-01-2015, 02:41 PM #26
- Join Date
- Dec 2000
- Location
- Bremerton WA USA
- Posts
- 9,413
Mark Rand has hit on the essential weakness of camelback straight edge design: they are not tortionally stiff. It would seem that a small twist could lurk in a camelback's reference face masked by its tortional compliance. I've never experimented to see if tortional stiffness is a real factor. Presuming the usual (excrutiating) care is taken in the averaging and equalling phases of the process, it would seem any twist will work out with the other errors.
The fact remains a traditional camelbacck straight edge can be represented as pretty much of an I beam: a flat plate (webbed tress actually) with a flange representing the reference edge and an elliptical section representing the bow. No great tortional stiffness there.
The much-admired Moore straight edge has a bulkheaded box section which, while intrinsically stiff, is bulky and does not lend itself as a scraping reference for long slender way bearings, narrow ledges, or to date has never been provided an extension intended to slip under the overhang of a dovetail.
Which returns us to the traditional camelback.
There is what might I call "lobing" but that suggests the wrong image. Maybe a better term would be "ruffling." If the necessary randomization is omitted, it is theoretically possible for a radialy symmetrical wave to develop that carried to extreme would resemble a circular ruffle. This is why it's reccommended to be a little sloppy when rotating and side shifting the work while generating plane reference surfaces by the Whitworth averaging of errors process. Randomness in the equalling phase works against order, regularity, and symmetry.
Maybe some hardy soul will experiment with his straight edge on his granite flat to see if tortional stiffness is a actually a factor and, if so, how to avoid it. -
Actually, NO. With such a small rotation, the two, essentially helical surfaces, will still maintain pretty good contact except at the center where they will separate. This separation at the center is a false indication and will lead to scraping the outer portions too much, but still with the same twist. So you will increase the error, not decrease it.
A 90 degree rotation is best to bring the opposite diagonals into coincidence as I show in the bottom of my drawing. And then, it becomes apparent that the narrow straight edges will not have enough area in contact to print more than a small percentage of their total area at one time AND the sensitivity of the test will be greatly reduced.
You really do need THREE plates and they really do need to be roughly square or round so they can be properly compared at 90 degree rotational increments, not long and skinny like a straight edge. Then use one of those flat plates to compare the straight edge to while scraping it.
Originally Posted by Peter.
Surely any twist would become apparent just by taking a print with the top straight edge rotated just enough so the left corner of the top straight edge was over the right corner of the bottom one? -
01-01-2015, 08:56 PM #28
- Join Date
- Jul 2007
- Location
- Rugby, Warwickshire. England
- Posts
- 3,103
I got caught by the helical straight edge problem, hence my comment up above. In the absence at the time, of a sufficiently large surface plate to check the straight edge, a level showed me the error of my ways (Pun intended).
-
01-01-2015, 10:58 PM #29
- Join Date
- Dec 2000
- Location
- Bremerton WA USA
- Posts
- 9,413
Man! I missed an important point. The hand of the postulated helix is rendered as a mirrored pair in the equaling phase of the process and mostly removed in the averaging phase. Subsequent cycles of averaging and equaling reduce the helix error to the minimum acheivable resolution.
Reply
- [machinist] Making Accurate Straight-Edges from Scratch by John A. Swensen - Page 2, Lawrence London, 01/02/2015
Archive powered by MHonArc 2.6.24.