[AustinT]

I have been thinking of how to build my next machine and have an idea. First i was thinking the best way to get the rails perfectly straight. Seems like a device could be built that could measure the deviation from a straight wire such as a piano wire stretched down the length of the machine. For example a linear potentiometer could be mounted on the linear block. The linear pot would need to have low friction so that the wire would not simply deflect. As the block was slid down the table the rail could be adjusted until it was straight.
I’m sure the idea above has been thought of and is probably widely used but lets take it a step further. Machines already exist that use feedback from a wire or string. One machine I can think of is a curb machine with slip forms. Instead of setting up forms and pouring them in a conventional manner, a string line is set up, when its time to pour the curb machine will follow the string line.
So how can we use this?
First we don’t want to have a piano wire stretched down our machines all the time, it needs to be a device that we set up and measure the deviation from straight linear motion. The deviation is recorded and used during normal operation of the machine.
For example let’s exaggerate and say that a certain location along the rail is out .125" from being straight. We know this because we have already set up our wire measuring system described earlier. During operation the machine would make up for this by superimposing the deviation with the planned route. On a gantry type machine if the x rails were not straight then the y direction steppers or servos would make up for it. If the y rails were not perfect then the x steppers or servos would make up for it.
Basically what it comes down to is open loop linear movement maintained by the software. CNC builders could hold tighter tolerances even if the rails were not set up perfect. They would however, need to make sure the rails were secure and had no slop.

[mxtras]

I think there are several ways to do what you are suggesting.

Here is one use of wire to achieve flatness: http://oneoceankayaks.com/madvac/madvac_index.htm

Scott

[mxtras]

I guess you are suggesting to use the wire as feedback rather than using it as an aid in mechanical construction?

Scott

[AustinT]

I guess you are suggesting to use the wire as feedback rather than using it as an aid in mechanical construction?

Scott

Hi,
Yes it would be feedback. The wire would only be used during the setup of the rail. The machine during setup the machine needs to know a linear distance down the rail and a deviation from the guide wire. This information is used later during the operation mode of the machine. Suppose a straight cut needs to be made down the x axis length of the machine. Traditionally we would only use the x axis servos or steppers to make the cut. Now we would know that during the cut certain locations along the rail are not straight and we would use the y axis steppers or servos to make up the difference.

[Geof]

.....For example let’s exaggerate and say that a certain location along the rail is out .125" from being straight.....They would however, need to make sure the rails were secure and had no slop.

If you have a deviation of 1/8" and no slop then something is going to bind very hard somewhere along the travel. You cannot have a rigid gantry structure ridng on a pair of rails that are not parallel and straight.

[mxtras]

I guess the way I look at it is this: The energy and effort spent to make such a system automatically compensate or correct for mechanical deficiencies would likely be better spent straightening out the mechanical deficiencies.

Scott

[AustinT]

If you have a deviation of 1/8" and no slop then something is going to bind very hard somewhere along the travel. You cannot have a rigid gantry structure ridng on a pair of rails that are not parallel and straight.

Hi, The .125" was too much of an exaggeration. When I built my plasma table I used two lasers, a dial indiator and a piece of cold rolled 3X .5" steel as a straight edge. One laser was a construction laser to give me elevation It has an accuracy of an 1/8" at 100' if I remember right. I was not too concerned about the elevation becasue the torch height control would make up for it later. The other laser was to help eliminate the rail from rolling or twisting. It mounted on the carriage on a swivel. I made a mark on a Door about 60 feet away. As I pushed the carriage down the track I made sure the laser stayed on the mark. The dial indicator was mounted on the carrige and the cold rold was secure to the table. I started at one end and used 10-32 set screws to adjust the linear rail until it was as straight as the cold rold, good enough for plasma cutting.
This was used as the master rail. The other rail on the opposite side of the table was adjusted so the gantry would run smooth without binding.
I guess what im getting at is both rails are ridgid and the gantry runs smooth. As an assembly however the whole ganty does not follow a perfectly linear path.
Again this is not a big deal on my plasma and I am happy with the way it came out. I am going to build a gantry style mill and would like to see how precise I can make the rails. The machine will be constructed using epoxy and granite. I like the low creep of the E/G and the idea of being able to pour a self leveing table.

[toastydeath]

Not to insult you, but I don't really think creep is a big concern for a home machine builder. People talk about creep after they're done designing the temperature and vibration control systems for the building they're going to put their new creep-sensitive million dollar machine tool in.

If you are going for accuracy, true accuracy, you need to focus more on what material is going to be the easiest for you to work and make corrections to with the tools at hand. Things like thermal expansion and creep are not even on the accuracy radar yet. And both of those are very complex issues that require an entirely separate discussion for each.

If that material turns out to be E/G, then go for it. But choose it because you can work it after it has been cast, not because it has properties that don't matter to someone without twenty or thirty thousand dollars in measurement equipment.

You can probably get your machine ways to an acceptable condition with a small machinist's level and a couple indicators, and some stock to make some measuring jigs. Check out Google for webpages and folks who do machine tool reconditioning who post their practices online, or alternatively, buy a book on the subject. There are a couple different roads to get you to the same endpoint.

One of the books people have always recommended to me, but I have not had the chance to read, is "Machine Tool Reconditioning" by Edward F. Connelly.

Additionally, some of the other forums online have pretty darn good information on getting a set of ways straight.

[AustinT]

I guess the way I look at it is this: The energy and effort spent to make such a system automatically compensate or correct for mechanical deficiencies would likely be better spent straightening out the mechanical deficiencies.

Scott

After thinking it over you are probably right it might not be worth the effort to make a system that automatically compensates for one machine . Would it be worth it however if it was standard input on a control such as Mach? It just seems like after it was built and working, Users could put down a couple of magnetic mounts with a wire stretched between the mounts . Then the machine in Mach could be slowly jogged from one end to the other. Mach would know the linear distance to the error and then would remember to compenstate for the error later on.

[AustinT]

Not to insult you, but I don't really think creep is a big concern for a home machine builder.

If you are going for accuracy, true accuracy, you need to focus more on what material is going to be the easiest for you to work and make corrections to with the tools at hand. .............If that material turns out to be E/G, then go for it. But choose it because you can work it after it has been cast, not because it has properties that don't matter to someone without twenty or thirty thousand dollars in measurement equipment.
......
One of the books people have always recommended to me, but I have not had the chance to read, is "Machine Tool Reconditioning" by Edward F. Connelly.

Additionally, some of the other forums online have pretty darn good information on getting a set of ways straight.

Thanks for the post, Im not sure at this point if E/G will be the easiest. I have some epoxy on its way so I can do some tests. The aggregate is already available as we own a ready mix company . The aggreate has a high percentage of quartz and i think it will work well. A lot of the concrete tools such as mixers, vibrators ect are already availabe as well.

Thanks for the recommended reading, I have been spending a good part of the day reasearching machine way scraping.

[toastydeath]

No problem. I also have a very heavy interest and career aspiration for precision machine tool building/rebuilding.

After thinking it over you are probably right it might not be worth the effort to make a system that automatically compensates for one machine . Would it be worth it however if it was standard input on a control such as Mach? It just seems like after it was built and working, Users could put down a couple of magnetic mounts with a wire stretched between the mounts . Then the machine in Mach could be slowly jogged from one end to the other. Mach would know the linear distance to the error and then would remember to compenstate for the error later on.

The problem I see with this is that error correction and compensation are only truly effective on a small (but relative) scale. A lot of machines I've seen on various forums built as a DIY project have larger, fundamental misalignment problems that make it hard to apply either correction or compensation.

A set of machine ways that are "fundamentally" misaligned will cause apparently unpredictable (nondeterministic, in the technical vocabulary) motions in the table that can't be properly compensated for. Poor ways on a linear axis will cause apparent unpredictability in not only the axis that has the error, but will cause linear and angular errors in the other axes that will change with load and other machine parameters.

When you get a set of guides, angular or linear, that true to a fine tolerance, error correction becomes more deterministic rather than apparently random. The errors in the axes are more repeatable due to the effects that take hold at higher tolerances (reliable elastic averaging) and become far more repeatable.

This is all under the dual assumption that there is a way to precisely align the string in the first place, and that there is a way to separate the vertical and horizontal components of the error motion.