PDA

View Full Version : TCT, RDM, test bar, etc.......

J Tiers
05-30-2012, 12:51 AM
Since the other thread got deservedly closed, for un-needed nastiness on several person's parts, including mine, here is some explanation of the TCT as I understand it from my reading and use. I'll be nice if you will....... 'nuff said

I will NOT call it "TCM", which would be the "two collars "method".....

NEITHER the Rollie's Dad's ****** NOR the two collars ******* should really be called a "method", in my biased opinion....

BOTH are "tests", giving "results".....

What you DO with those test results is your own business. I do not believe it is possible to generalize on what to DO with the results, because machines vary, and what is true of one may not be true of another.

One machine may be worn in such a way that shimming and twisting into a semblance of straightness is not possible, because the leveling surfaces are not aligned with the slideway surfaces.

The slideways may be worn such that both front and rear slope down towards the headstock, and the front is worn more, adding a "twist", the total perhaps being as much as 15 thou or even more wear (I have seen that).

Without a geometrical proof, my speculation is that there is no practical way to re-align by twisting and shimming to remove that problem, since there will always remain a "knuckle" in front of the headstock where the slideway starts to slope up towards the tailstock. A reasonable average may be as good as it gets....and that in turn may cause issues with the contact of carriage to slideway as it moves along the twisted bed.

Let that be an argument for later.......

OK.... so my point is that conflating "the test which reveals the fault" and "the solution to the fault" is where RDM and other "methods" or "systems" fall out of touch with reality.

lets stick to the "test" part of each.....

***********

1) RDM test..... this is a test using an arbitrary bar mounted in an arbitrary manner to the spindle. Let's say it is chucked, but it could be in a collet, it could have a taper end, etc.

The test is done by rotating the bar (and spindle), with bar in contact with an indicator secured to the bed (probably on the carriage), noting the high and low readings, and averaging to get a single number. This is done at say two places along the bed, and the result is to arrive at a reasonably accurate relation between the spindle axis and the bed. Usually horizontal measurements are of most interest, but the vertical is also measurable.

Error sources include indicator repeatability, flex in the bar (they are typically but not required to be, thin), errors reading the indicator each time, roundness of the bar (affects true center) and error of bar straightness, which may force use of a less sensitive but longer "throw" indicator, plus errors in finding the true max and min readings. I am sure there are numerous others (mostly smaller) as well.

Typically, the results of this are used to advise on shimming of the bed, or headstock adjustments to improve alignment. I don't propose to get involved with this aspect. I would advise leveling the lathe first, but many of the RDM descriptions tout the "method" and suggest that it will replace the level. I dispute that, and I have company......

net result is that you get numbers relating to the alignment of the H/S spindle axis with the *effective* bed (including all wear)

***************

Test bar........... First you level the lathe. Then a precision bar with proper taper end is inserted in the spindle taper, and the indicator is traversed along it on the carriage to assess the error of carriage movement vs spindle center.

Since things are rarely perfect, a variety of the same averaging routine may need to be done to arrive at a proper assessment. This becomes the same thing as RDM, only the bar costs more.

Naturally, if you are the manufacturer, things are different......The socket is being tested as much as the alignment.... wobble out of spec means the spindle socket is bad, and needs re-working. But for the rest of us, it is RDM with expensive parts.

**********

TCM (Forrest's)...... This *new innovation* is essentially the same thing as having a test bar, except that you can MAKE the bar, the key being identical diameters of two (or three etc) collars which you turn to suit, after which you use the bar as-is. You would naturally level the machine first.

As with the test bar and RDM, averaging may be required. Again, if so, same thing as RDM, but more trouble to make the bar.....

*************

TCT......... In this test, you again hold a bar (or tubular piece) in a chuck or any other convenient holder on the spindle. (See any machining text for a description of this test, BTW)

The first time of use, you turn it to a "dogbone" shape with two (or 3) collars on it as with Forrest's test. However, instead of checking with indicators, you actually take a slow, fine, cut, with power feed, across both collars, using a tool (and bar material) that will give best finish possible. When you have completed the pass, and a full cut was taken on both collars (no "winking", or spots not cut), you measure the diameters with your best mic.

If they are the same, you win... your machine turns spot on over that distance, and the H/S is aligned with the slideways sufficiently to do that.

if the T/S end is larger, either the H/S is aimed towards the back, away from the tool path, OR the bar flexed, OR there is a gross vertical misalignment, OR there is a bed wear issue. You can check the vertical alinement, and you can use a large not-very-flexible bar or tube, which leaves the aiming, or a bedwear issue.

if the H/S end is larger, the H/S is aimed towards the FRONT, or else one of the other issues is at fault, with the same remedies.

Error sources include accuracy of turning (at least partly a feature of the machine), surface finish of the cut, and inconsistency of the mic or the user's technique.

These can be controlled to some degree, plus they will also affect your accurate turning results, so they are pretty much a 'fact" of the machine..... What you get is what you will get in using the machine..... complete with all it's "assignable causes of variation" and limits of accuracy. It is an "honest test".

However, there is only ONE measurement at each point, not two..... half the chances for errors, easier to be careful. And, the check is a comparison..... no absolute measurement, and the mic only needs to have the accuracy of resolution plus repeatability, doesn't even need to be calibrated.

This is the classic "two collars test" as described in many machining texts.

A short test bar (tube) chucking end at left. As you may notice, it's been used some, the collars are down a bit.........

http://img.photobucket.com/albums/0803/jstanley/Testbar.jpg

05-30-2012, 01:40 PM
For reference, my position on RDM is copied below. From what you've written we may agree on more things than either of us realized. Separating the measurement method from the correction method is a major step because Wasser's inclusion of it confused the issue immensely.

Clearly, there are differences remaining so I'll make a subsequent post with questions so you can clarify areas I don't understand.

John

----------------------------------------------

My view is all three headstock/bed alignment methods are in essence the same: if you have a precision test bar then you use it. If you don't have a precision test bar, you simulate it. With TCT you turn two accurate test areas on a bar and use them to simulate the precision test bar. With RDM you remove the error sources using arithmetic thus simulating a precision test bar. However it's done, the same information must be extracted from the test used. The goal remains: measure the deviation between spindle axis and bed ways.

I consider John Wasser's paper on RDM to be a concept paper rather than a test plan ready for execution. The paper identifies error sources that make chucking a generic test bar differ from measuring a precision test bar fitted in the headstock taper. And he indicates how these errors could be compensated for mathematically. Finally, he describes how the results might be used to improve the situation.

In reducing the concept to practice, one need not (in my view should not) simply follow what he wrote. I believe the reason the error sources are identified is so the user can figure a way to minimize them. For example, providing a method to allow using a rod whose diameter isn't constant and removing the difference mathematically would be plain silly - I think it is included so people understand that rod diameter is critical to the measurement. Using an accurately ground rod with a major bend in it would be silly too. But, if the rod used has a slight bend the calculation should prevent it from affecting the result, where his logic on this seems reasonable.

The test bar will have some runout from chucking which is unavoidable. Runout at the far end can be minimized easily by snugging the chuck in increments, measuring the high point and pushing it down with your thumb, tightening more, etc. It's easy to get TIR within a thou at the far end. This makes it more closely simulate the perfect test bar we wish we had, making the mathematical corrections needed smaller. Then, set the indicator up for horizontal or vertical measurement and set indicator zero so the reading for runout at the chuck end goes equally above and below zero, making John's Near End Average = 0 so it can be ignored in the calculation. Now, the RDM calculation is reduced to adding the high and low readings at the far end together and dividing by 2. (This result is analogous to the TCT result except with TCT the result reads directly off the indicator since runout has been removed by machining.)

A detail John's paper ignored is test bar sag. All test bars sag, even precision test bars. A larger diameter test bar sags less. In all of the methods sag should be accounted for. Measuring the actual sag isn't easy so generally it is calculated and then used to adjust the vertical result.

John Wasser's paper suggests shimming the feet of the lathe to correct bed twist. He didn't say to check your common sense or experience at the door. It is reasonable to use whaterver means are available to ensure the bed isn't twisted before using any spindle/bed alignment method. Nor does his paper consider the effect of bed wear on the measurement so common sense and experience are needed there too.

Does RDM provide perfect results? Certainly not - although it looks like you're taking only two measurements per axis, setting indicator zero at the chuck end is really 2 more readings that enter into the calculation invisibly. So now there are 4 readings, each with a little error and if the errors don't cancel then it would be easy to be off half a thou. The ground test bar could have a diameter difference between the measurement points of a tenth since I can't measure better than that. The calculated sag in the test bar gets added in too and it's a calculated value which could be affected by the type of steel in the test bar. And so it goes -there is some noise in the result. It's random so sometimes it cancels and sometimes it adds. The question is, is the error in RDM greater than in other methods. And does it matter in a home shop - that is, there are some real world limitations imposed by Mother Nature when you try to use the results to adjust a lathe.

lazlo
05-30-2012, 01:59 PM
From what you've written we may agree on more things than either of us realized.

I've found that's often the case Jan ;)

lakeside53
05-30-2012, 02:32 PM
Or as is often (sadly) the case "we agree on more than we'd like to admit";)

05-30-2012, 02:44 PM
In reading what you've written, you agree that the measurement method and the use of the results are separate issues and what we're considering is simply the measurement aspect.

Further, you seem agree that all three home shop methods: TCT, TCM, and RDM can use whatever size test bar is convenient so sag is equivalent for all.

I am impressed that you said the test bar method and TCM are really RDM in disguise - I speculated that this was the case but wasn't sure I could defend that so I didn't raise that issue.

The major issue seems to be consideration of measurement noise and how it affects the calculated error value in the real world.

If there are other areas where we disagree it would be helpful if you could identify them at this point.

John

Edit: change "measured value" to "calculated error value"

dian
05-30-2012, 03:51 PM
cutting the barbell works for me. i have a short and sturdy lathe.

however, i can see problems with it. firstly cutting pressure. i suspect it will always be larger than the force of an indicator. thus, the lenghth you can check is limited. my largest micrometer is 75 mm, so i am pretty much limited to 300 mm. yes, i could chuck up a 150 mm ally pipe and have it stick out 600 mm, but i would have to get the pipe and the micrometer first.

secondly it depends on what you want to do. if you want to make sure, your chucked work comes out straight, its fine. when taking off material, you will have a much smaller stickout. but lets suppose you have a long bed and want to make an assesment on its wear. you want to check it in 10 locations over 1500 mm lets say. then i agree, that the "measurement approaches" might work better.

has anybody tried to elliminate sag by supporting the bar by a spring or piece of rubber?

test bars: my understanding is, that "test bars" are ground to a tolerance of 0.005 mm, whatever that exactly means. i also understand, a mt3/300 mm bar will set me back 300\$. ( i would need a mt5, dont even want to think about cost.) also it will be only as good as the spindle taper. straight bars are out of the question, because of the runout of the chuck or collet. btw, i have four test bars, one even has a certificate and none of them are true, especially since i have a miutoyo 0.001.

idea: hard turn the jaws (or get soft jaws) for the diameter of the test bar.

gadget, the sag will not depent on the type of steel, i believe it will even not be substantially be influenced if you use steel or ally, as the weight is roughly proportional to youngs modulus.

Forestgnome
05-30-2012, 04:46 PM
Sag wouldn't generally be a factor unless you're testing over a great length. For instance, a 1" diameter aluminum bar 12" long would sag 4.5e-7 inches at the unsupported end. The same size steel bar would be 1.6e-7 inches. I would just be careful of how large you make the "dumbell". It shouldn't be much larger than the minor diameter. If you wanted to go extreme I suppose you could press fit aluminum dumbells onto a steel shaft.

lazlo
05-30-2012, 05:13 PM
gadget, the sag will not depent on the type of steel, i believe it will even not be substantially be influenced if you use steel or ally, as the weight is roughly proportional to youngs modulus.

+1. In Harry Beckley's outstanding "Wreck" rebuild, he calculates the sag for his TCT test bar. It wasn't insignificant.

Connelly, IIRC, recommends a hollow tube, which will have less sag. A tapered tube would be even better, if you can calculate the moment correctly.

beckley23
05-30-2012, 06:14 PM
Figuring the sag starts at post #126 in the "Another New Toy" topic on PM.
http://www.practicalmachinist.com/vb/monarch-lathes/another-new-toy-163406/index4.html
I had some help with the numbers.
Harry

J Tiers
05-30-2012, 11:33 PM
To clarify....

Test bar, TCM, RDM are, or may be, essentially similar. TCT (distinct from TCM) is different.

Actually, the test bar is also different. I say this because the test bar, as commonly used, has the spindle taper on it, and is always used placed into the spindle.

Therefore, it is really testing the accuracy of the spindle taper, in terms of centering, as well as concentricity.... i.e. absolute overall off-center, AND "off center aim".
The bar may be perfectly parallel to the rotational center, but offset from it, so the whole bar is moving identically off-center. The bar may also be set at an angle to the rotational center, such that the far end is "out" more than the near end.

The test bar can*also* be used to check aim, in which case your errors mean you must treat it as the bar in RDM is treated, averaging errors.

Now, RDM is no new innovation in measurement..... In truth, while I am separating the *test* from the *correction method*, this is NOT the standard case with RDM....they are presented as a package, and touted as a "method" for alignment needing no level. However, the *measurement* part of RDM uses standard methods which pre-date the emergence of RDM itself.

Therefore it is more correct to suggest that RDM uses the same techniques as the test bar would for very precise measuring, rather than suggesting that the bar test is based on RDM.

The key issue in TCT vs TCM (Forrest's method) is that you have no standard "test bar", you use the bar only to test the alignment by actually making a cut. In contrast, Forrest suggests making the bar, and then using it as a standard, which simply shifts it over to being a "test bar" type measurement.

Issues:

before discussing ANY of the issues raised, note that one is not (at least I am not) trying to make a full-bed survey with the test. The most precise printed guarantee I have seen given for a lathe is the 0.0001" in six inches given in the specifications for the Rivett 608. Note that is a "turning" guarantee, i.e. making a cut, not using an indicator on a test bar.

One presumes that this is primarily for the first six inches away from the headstock, although I suppose nothing precludes it from being for ANY six inch length along the bed.... So presumably, this was tested with a cut as per TCT, with the test bar being about 6 inches long outside the spindle. Assuming the bed is straight, that simply comes down in the first place to the sum of the "aiming error" PLUS the bed straightness error being no more than 0.0001" over 6 inches.

Don't try to make the test more than it is, or can be.

So deflection and sag were raised as problems.......

***Deflection....

Deflection of the test bar is handled by making it bigger in diameter ....and using a very fine low-pressure cut, so that it does not deflect enough to be a problem. Considering the construction of the machine, this should be possible to reduce to an amount less than exists elsewhere in the structure, by upsizing bar diameter.

Deflections in parts OTHER than the bar itself are *fair game*, a part of the machine, and form the inevitable "noise floor" that stops you from getting more precise in your measurements, OR your work, with that machine.

Trying to exceed the noise floor is not productive...... which is why RDM manages to work out decently... most machines it is used on have problems and issues that raise the noise floor enough to swamp much of the measurement error sources in RDM.
.
.

***Sag in the bar......

This is again handled by a larger bar, so that it sags very little. Also the larger bar has less diameter difference resulting from a constant sag, because the periphery is closer to being a vertical plane at the cutter, the larger the bar diameter is. The diameter difference per thou of up or down error is vanishingly low.

While sag is not irrelevant, it is, with a large bar and a shorter distance, a secondary effect. If you want to go out a half meter with an unsupported bar, well, you will need to deal with it, but that is extreme. I don't suppose any cutting-type test is useful with that amount of stickout on any lathe most of us will have or want.

You will need to use a test bar of some sort to do that type extreme length test..... and you should expect to deal with a larger absolute error than over a shorter distance. An error of 0.001" at 20" is, if it is smooth, the same as an error of 0.00025" over 5", and is nothing to sneeze at.

dian
05-31-2012, 04:43 AM
forest, can you explain your calculations? i wonder, because last time i had a look at the formula, i saw density in the first power above the line and the modulus in the first power below the line. since aluminum has a modulus and density that are roughly a third of the value for steel, i figured deflection would be similar for both.

Forestgnome
05-31-2012, 10:21 AM
Figuring the sag starts at post #126 in the "Another New Toy" topic on PM.
http://www.practicalmachinist.com/vb/monarch-lathes/another-new-toy-163406/index4.html
I had some help with the numbers.
Harry

I get 4.9e-7 inches in the case you referred to. Quite a bit different than .0007". I'm using ml^2/3 for the moment and ml^3/(8EI) for the deflection. Specified rod was 1018 1.34"x15".

lazlo
05-31-2012, 11:36 AM
***Sag in the bar......

This is again handled by a larger bar, so that it sags very little.

A larger bar will sag more. The two things you can do to reduce sag is to reduce the length/diameter ratio, but then you're losing resolution on the two collars test, or use a hollow test bar, with a tapered hollow having the least sag of all.

Connelly has a whole chapter on making and measuring test bars in Chapter 15, starting on page 109. It's a really good read, in a very dry book :)

lazlo
05-31-2012, 11:43 AM
I get 4.9e-7 inches in the case you referred to. Quite a bit different than .0007". I'm using ml^2/3 for the moment and ml^3/(8EI) for the deflection. Specified rod was 1018 1.34"x15".

I'm getting 5 tenths -- what DaveE got. Are you using PiD^4/64 for the circular moment?
I'm using the beam calculation from Machinery's Handbook...

Forestgnome
05-31-2012, 11:55 AM
I'm getting 5 tenths -- what DaveE got. Are you using PiD^4/64 for the circular moment?
I'm using the beam calculation from Machinery's Handbook...

I could certainly be using the formulas wrong, but I'm using ml^2/3 for the moment. That's the moment for a slender rod. I'm using the deflection formula for a uniformly distributed load on a cantilevered beam.

Forestgnome
05-31-2012, 12:01 PM
I do believe I was using the wrong moment. I'm not an engineer. When I was looking for a moment for a rod, the formula I found didn't quite make sense for the application. I needed the area moment of inertia, which is what you provided. Thanks for the correction.

jep24601
05-31-2012, 01:06 PM
Since you guys are calculating to fractions of a thou you should be aware that most textbook calculations only calculate for bending deflection and not shear deflection. If you want to be extremely accurate you have to add the shear deflection to the bending deflection to get the true deflection. However, shear deflection is small compared to bending deflection in most cases.

dian
05-31-2012, 01:34 PM

Forestgnome
05-31-2012, 03:04 PM
I found the calculations for shear deflection, but no info on what the form factor (alpha) would be for a circular section.

05-31-2012, 03:22 PM
I note you edited your description of TCT execution, and this is fine with me. However, since it is your stake in the ground, it would be helpful if you add a note to your latest post at the time you make an edit, indicating roughly what you changed so I know what to look for. By the same token, if I need to change my description of RDM I'll advise you.

I'm using the prior thread for context to help in understanding the issues you have with RDM. The last thing you said there was:
"I believe that it has more measurements and settings, and is a "one step removed" test, causing it to have more error sources than the "re-cut when used" TCT/TCM."

In your latest post you allude to "Any of the issues raised". Could you enumerate and detail the issues you're referring to? If you can break these into components which we can go through one at a time that would make it easier to deal with.

I assure you I have no intention of expanding our discussion into a full bed survey - my perspective is that we're discussing the very narrow area of how RDM and TCT compare for use in the home shop. If we stick to this narrow area then my thought on an agenda, once you provide a list of the issues you have, would be to discuss: 1) measurements and the differences in measurements between the two approaches, 2) details involved in turning the two collars and their effect on measurements, 3) compare applying the two approaches in the home shop while avoiding precisely how corrections are made. If we get through that without bloodshed, I have a couple speculations about RDM and what it can tell the user that I'd like an opinion on.

In line with maintaining a narrow focus, I have agreed to discuss RDM execution per the description I provided and compare it to TCT execution per the description you provided, where both are subject to real world conditions. To this end, I'd like to exclude discussion of other versions of RDM which conflate measurement and correction.

------------------------------------------

If you have questions or issues with my description of RDM, this would be a good time to raise them and I'll modify the description as needed.

In a similar vein, in your opening post you say:

"Error sources include indicator repeatability, flex in the bar (they are typically but not required to be, thin), errors reading the indicator each time, roundness of the bar (affects true center) and error of bar straightness, which may force use of a less sensitive but longer "throw" indicator, plus errors in finding the true max and min readings. I am sure there are numerous others (mostly smaller) as well."

You indicate that errors in bar straightness may require using a less sensitive indicator. In my description I say that maximum runout is easily set to 1 thou or so. I also specify a bar with minimal kink. Under these conditions, I believe the RDM calculation removes any issues with bar straightness. Could you expand on why the choice of indicator is affected?

In your description of TCT you don't specify how the tool center height is set. I believe this matters if you intend to deal with large vertical errors, as you specify. While it is a cosine type error, for large vertical errors shouldn't it be considered, if only to show that it can be ignored?

In your description of TCT you mention large errors in height and/or front-back but don't describe how you get from large errors to small errors where (I assume) mic readings are all that is needed. One of the claims against RDM is its complexity. Part of this complexity involves measuring both H and V errors, which RDM does for large or small errors. In order to compare RDM and TCT as to complexity, you need to show how H and V are separated in order to get to the small error case,
You can see how this is handled with my version of RDM, in detail, by visiting my site:
While it could be added to my description here, it would needlessly clutter it. I will respond to questions about that description as if they are part of the concise description here.

In the prior thread a criticism you leveled was that RDM was difficult for a new user to follow. It seems fair to ask that your description of TCT be as easy to follow as my description of RDM. I can assure you that a fair number of new users have followed my RDM description - so far, only the happy ones have written to me ;-)

I will admit to an ulterior motive here. I have not found an execution plan for TCT on the web, let alone one that new users could execute. If you come up with one and can show me that it equals RDM then I can link to this thread as an alternative to RDM - currently I link to the J. Latta paper. Your little survey of the various methods is a fringe benefit.

Forestgnome
05-31-2012, 06:21 PM
Found a reference to the shear deflection constant for a circular section as10/9. My calculation for the 1.34"x15" rod would be 5e-5" for the shear deflection, which is added to the bending deflection. Does this sound right?

philbur
05-31-2012, 07:08 PM
I think you are deep in the long grass. Here's a quote:

"Deflections due to Shear: Generally deflections due to shear can be neglected as small (<1%) compared to deflections due to moments. However, for short, heavily-loaded beams, this deflection can be significant and an approximate method can be used to evaluate it."

Reference page 3.16 at: http://courses.washington.edu/me354a/chap3.pdf

If you need to do deflection calculations stick with simple bending deflection.

Phil:)

Found a reference to the shear deflection constant for a circular section as10/9. My calculation for the 1.34"x15" rod would be 5e-5" for the shear deflection, which is added to the bending deflection. Does this sound right?

jep24601
05-31-2012, 09:30 PM
http://i172.photobucket.com/albums/w23/jep24601/Shop/ShearDef.gif

J Tiers
05-31-2012, 10:44 PM

I don't know what I "edited", I have been consistently describing the exact same procedure..... or at least trying to. The definition I am using is not, as you seem to again imply, "shifting or squirming" to meet circumstances. That background "basso continuo" is a bit puzzling...... You have said you are not a machinist (an obvious fib, by the way), so I may as well say that I am not an academic, which may be a somewhat equivalent fib, but serves to point out that I am not nit picking the last details. I have quite enough of that to do in "real life" and prefer to avoid it here.

The TCT I suggest is for exactly ONE purpose, and ONE purpose only....

That is to assess the degree of taper turning error that a given machine has, when turning a workpiece supported only by the headstock.

That is the basic issue that is the most bothersome and problematic for most users, and is the reason for the emergence of RDM etc.

There is no desire on my part to extend the test to vertical issues.... in the first place, they are much less of an issue for most lathe usage, since the vertical alignment has much less effect on diameter for the most common ranges of work. For very fine work on small diameter work, there may be an issue, but it is worth noting that small diameter work is not commonly extended out several inches, so the absolute effects are reduced.

However, once the collars are equal, one can easily run an indicator over the tops if you are curious. That would be basically a test bar, except that there should be no need for double measurements and averaging, a straight comparison should be accurate enough to equal the machine accuracy, the collars are already as accurately concentric as that machine will make them.

As for the height of tool, that should be obvious as it is the same as for any other turning..... on center.

I am not interested in arguing the last few Scovill points about this hot sauce... It is what it is, and I submit that it is more direct and likely even easier than the various indicator methods of doing the same thing. I also claim it to be "honest", not offering the false promise of better accuracy than the machine in question can truly produce.

I claim it is direct because it uses the machine as it is intended to be used....to cut metal, which necessarily produces a result reflecting the machine as it is, with all faults.

I claim it as easier, because it only requires you to chuck and cut metal, and measure the results. There is no taking of averages, calculating of sags, no zeroing at one end, no adjusting for best bar alinement, etc, etc.

I claim it as "honest" because again it is a practical test, using the machine as it will be used, employing all the parts which produce inaccuracy, although in a way intended to give the very best results possible. For that reason it is more truly representing the best you can expect, and that "best" is one that you know you CAN get, because you just DID get it.

For a proof of adherence to the guarantee of turning accuracy.... I would far prefer a test sheet in which the diameters of the two (or three) collars which were actually turned on the machine are written on the test sheet. That says "we checked it and it turned the test piece to the specified and guaranteed accuracy just as we promise". Full stop, nothing to explain.

I doubt if most practical folks would be as impressed by a test sheet covered with arithmetic calculations showing that the mean position of the test bar was within XX of parallel to the path of the carriage as read from the indicator...........

My claim is that all the various measuring and so forth may be able to get a very precise result, but you have far less assurance that you can GET that result when using the machine. After all, you have eliminated some normally present factors, and can only assume that the machine will still be that accurate with cutting pressure present.

And you are still left with roundness errors of the bar..... one thing it needs to be is round.... even if not straight. otherwise if it WAS straight, and perfectly aligned with the spindle axis, it would STILL show variations.... So the whole argument about the usability of an imperfect bar tend to fall down.... all it says is that in effect you can use a cast-off bent precision test bar .......;)

With TCT you MAKE it as round as the machine can. If the machine cannot make it rounder, you have reached the limit beyond which you cannot go, so no point in attempting tighter measurements.

As for the indicator...... most tenths indicators have limited range of motion.... if the imperfect test bar is too bad, you exceed that range and must use a longer travel, but lower resolution indicator. The other choice may be to do a lot of adjustment of the bar to minimize the error, in which case the "complication" argument may rapidly get to a fairly obvious "game over" situation.

Now I will violate the terms of our "agreement", at least the ones you set forth.............with a complete lack of remorse, I am afraid.......

My MAJOR issue with RDM as it is presented to the great unwashed, is that it generally claims to be "one stop shopping" conflating a way of measuring headstock misalinement, a way of measuring the effect of wear, AND a way of measuring bed twist, in a single measurement..... This is a patent falsehood, as you have two (or three) unknowns and only one value.

The only way that can "work" is if you take a "big picture" approach, and say "I don't give a rip WHY it isn't turning straight, all I want is to make it cut without taper".

In so doing, you must attempt to fix all ills with bed twisting, *as measured by the bar and indicator*. It seems clear that there may be one best 'countertwist" that provides a "best fit to the curve", but it is questionable whether one can truly get there in a sensible time, and without attempting the impossible...... Particularly if one is not that familiar with lathes and machine work.

Mind, I don't see TCT as any better for that..... I say the attempt is likely doomed to begin with, and NO measurement method can save it. The only hope one can have of a good result is if the headstock is alined OK, the effect of wear is minimal, and all you have to contend with is twist.

The *method of measuring* presented in RDM is "OK"..... It is intended to use available materials, and is limited by them. But it works to its limits, even if I think it is more complicated than the TCT, less representative of what is truly possible with the machine, and potentially prone to more errors.

A larger bar will sag more. The two things you can do to reduce sag is to reduce the length/diameter ratio, but then you're losing resolution on the two collars test, or use a hollow test bar, with a tapered hollow having the least sag of all.

If you leave the length the same, and increase diameter, it would appear that you have changed the L/D ratio..... So how does it sag MORE when I say it, but LESS when YOU say it?;)

You can compare the stiffness increase as diameter increases, with the mass increase, and I think, if geometry has not been as badly re-written as recent history has, that stiffness still gets larger faster....

As for the resolution, I don't see that it is changed at all, unless your idea is to take a hacksaw to the bar.........

Perhaps you had some other idea of "bigger"?

You will note that the bar I illustrated is hollow.

beckley23
05-31-2012, 10:48 PM
Gentlemen,
This discussion you're having about the sag, and how to figure it, of the test bar is pointless. The question and answer, from 2 engineers, was posted 3 years ago. I ran with the number they provided, and the tailstock came out just fine. I used the number they provided, added .0005" to it, for the maximum high point at the end of the bar, so I could scrape the the TS, and achieve the tolerance on the test card.
That topic has had over 50,000 views, and surely had Russ and Dave made an error in their calculations, they would have been called on it. If you read further into the topic, you will see that there is not one contradictory comment regarding the sag calculation. Close to the end of the topic, are the final test cuts. In the short time that I owned the lathe, I never had a tolerance problem, and I never had to fight the lathe to hold a .0005" unilateral tolerance., and considering what I did to the lathe wasn't a full reconditioning, my expectations were more than fulfilled.
Harry

05-31-2012, 11:20 PM
Wow! 25 posts in 48 hours. We must be having fun.

I don;t know if it means anything to the current discussion but Jerry and I have slightly different approaches to the TCM:

Jerry turns the collars at the same tool setting and mikes the diference

I turn both collars to the same diameter (separately if need be) and scan them with a DTI.

You might think what's the difference? There is a small difference. Turning the collars at the same setting demonstrates straight turning. Turning both collars to the same diameter and scanning them demonstrates carriage tracking parallelism to the spindle axis.

My reasoning is a straight turning test may validate compensating errors: for example slight taper and tool wear. Scanning separates the readings from the turning test.

There's no reason a DTI cannot be used in Jerry's version - post cutting test - provided differences in collar diameter are compensated for in the results.

You need to set a DTI to register in the vertical plane as well to determine axis rise or droop towards the tailstock. I seem to recall spindle axis 0.0004" per ft rise towards the tailstock is allowed no droop

dp
05-31-2012, 11:24 PM
You might think what's the difference?

I think the principle difference is Jerry is turning a single but interrupted taper, and you are turning two independent tapers. I agree it doesn't matter much.

lazlo
05-31-2012, 11:41 PM
If you leave the length the same, and increase diameter, it would appear that you have changed the L/D ratio.....

My point is that the sag is proportional to the Length/Width ratio: the area moment is proportional to Diameter^4, and the sag is proportional to Length^4.

Axis on the right is sag in inches. So Harry would only improve his sag by about a tenth by going to a 1.5" bar, about 2 tenths going to a 1 3/4" bar.

http://i164.photobucket.com/albums/u15/rtgeorge_album/sag.jpg

So rather than increase the diameter, which gives you a very small improvement in sag for a massive increase in weight, it's much more effective to use a hollow cylinder, which is much stiffer per Length/Diameter ratio.

Which is, not surprisingly, why Connelly recommends using a hollow test bar :)

lazlo
05-31-2012, 11:48 PM
This discussion you're having about the sag, and how to figure it, of the test bar is pointless.

I think we all agree on how to calculate the sag. My only point was that Connelly recommends a hollow TCT bar to reduce sag.

J Tiers
06-01-2012, 12:21 AM
Which is, not surprisingly, why Connelly recommends using a hollow test bar :)

And even less surprisingly, why I actually USE a hollow "bar", and have suggested it right in these discussions..... ;) :p

I don;t know if it means anything to the current discussion but Jerry and I have slightly different approaches to the TCM:

Jerry turns the collars at the same tool setting and mikes the diference

I turn both collars to the same diameter (separately if need be) and scan them with a DTI.

My reasoning is a straight turning test may validate compensating errors: for example slight taper and tool wear for example. Scanning separates the readings from the turning test.

Yeah, that's why I call yours the "TCM" (might also be "JLATB"), and the textbook version (which I describe) the "TCT".....

It ain't just me using it..... it's been a standard test.... described in several textbooks and practical manuals I have..... I surely didn't invent it any more than Rollie's Dad invented his system.

perfect? likely not. Very few things are perfect.

Compensating errors? Could happen. Can still happen using a DTI with the "just like a test bar" (JLATB) method of same size collars turned first...... You may have a misalignment compensated by a wear pattern..... it takes outside tests and setup to eliminate the other factors....

Remember, one "known" has to be available for every separate equation save one...... if you want the headstock alinement to be tested, you have to already KNOW the machine is level, and how much wear it has, so you can eliminate those effects. That's the downfall of RDM.... not enough "knowns" to cover the "equations".

Anyhow, one presumes that the tester will have leveled the machine, and checked other things first..... What's the point of a precision test with a machine that has other errors of 10x the precision of your measurement? Anything you find will be "lost in the noise".

philbur
06-01-2012, 03:37 AM
This refers to a beam simply supported at both ends. I think with a cantilever the deflection due to bending is proportionally much greater.

Phil:)

http://i172.photobucket.com/albums/w23/jep24601/Shop/ShearDef.gif

06-01-2012, 03:40 AM
One thing not mentioned that affects the readings obtained by this test is tracking error in the saddle. If the bedways in the saddle and the ways on the bed are worn its conceivable that the carriage may track funny. Carraige reversal may cause the saddle to slew on the yaw axis, drift laterally, rock on the pitch axis, or a combination.

There are three degrees of freedon for tranlastion and rotation. In a new machine these degrees have been carefully worked to near zero except for the sole degree of freedom for which that axis is intended. Wear may degrade a motion that was once linear to one having components in any or all of the five remaining degrees of freedom.

This is not a theorectical exercise to keep the mental muscles pumped. One of my favorite tests for lathe condition is to place an indicator on the carriage to register its vertical movement to the bed. Then I apply a downward force on the end of the saddle wing over the way, a force equal to about 1/4 the weight of the carriage. On a lathe in like new condition there will be nearly zero change in the level. There will be some movement in a machine less than perfect and on down to a lathe in poor condition where the DTI spins 'round like the button on a \$hit house door.

I have climbed on the 3 f x 5 ft table of a relic HBM and heard it slurp air under the ways as the table rocked. These tenths chasing theoretical evolutions have a way of meeting blunt reality in the form of older machine tools in different stages of decrepitude.

darryl
06-01-2012, 04:42 AM
Two things- Forrest has brought up one that is pretty easy to check. With a carriage-mounted indicator and pretty much any smooth bar mounted in the headstock, you can check for unwanted motions in the carriage to bed interface, and you can check that anywhere along the bed. The actual readings don't matter, only deviations that seem excessive and/or don't return to zero when you try to move the carriage around. This seems like a good way to see the effect of worn ways near the H/S., and also to characterize the behavior of the carriage over the unworn parts of the bed. You would certainly want to deduce some intelligence from this before you go about trusting some readings you're getting from any of the methods that have been talked about.

Secondly- this is more of a question- in the opinion of the most experienced here, where is the most likely area where a turning error would derive from- worn ways, twisted bed, or mis-aligned headstock- or other? Now I'm coming from the 'what can I do about this' part of the alignment procedure. First is getting an indication that there is an error, second might be measuring the degree of the error, third and quite importantly might be disclosing the source or sources of the error or errors, then and only then does the fixing it come in.

I haven't seen much said about zeroing in on the actual source or sources of error.

oldtiffie
06-01-2012, 05:26 AM
http://i172.photobucket.com/albums/w23/jep24601/Shop/ShearDef.gif

This refers to a beam simply supported at both ends. I think with a cantilever the deflection due to bending is proportionally much greater.

Phil:)

The computation refers to a simply loaded beam - ie supported at each end.

The test bar (any of the methods discussed) is a cantilever suported only at one end.

That one end (support) in the chuck will be right at the front face of the chuck jaws but if the jaws are bell-mouthed the support may in fact be further back from the front face.

The properties of cantilever beams (neglecting shear) is at page 265 of Machinery's Hand Book 27 under "Beam Stress and Defelection Tables".

The general case for deflection of simple constant section cantilever beams is: (Wl^3)/(3EI).

I will leave it to others to work the deflection out for specific cases.

This may help:

06-01-2012, 06:03 AM
...this is more of a question- in the opinion of the most experienced here, where is the most likely area where a turning error would derive from- worn ways, twisted bed, or mis-aligned headstock- or other? Now I'm coming from the 'what can I do about this' part of the alignment procedure. First is getting an indication that there is an error, second might be measuring the degree of the error, third and quite importantly might be disclosing the source or sources of the error or errors, then and only then does the fixing it come in.
.

I don't think I ever ran a lathe I could trust to go direct to size simply by dialing referencing from some other feature even when I've set the DRO. There are variables conected to tooling, feeds and speeds, material consistancy etc that limit going directly to size without interim measurements. Most of the time, yeah, pretty much but whenever I shoot for finer tolerances I almost always stop a couple of cuts short of a +0.0005"/-0.0000" tolerance and sneak up taking half the first cut then mike and dial in for an adjusted second half holding back a little to stone or dress into the top of the tol. Most of the time this takes but a few extra minutes even on a clunky worn lathe or mill.

My home shop lathe is a dandy. I would say it's the equal of a Monarch EE (swagger! Strut!) for size holding if not for sensitivity and spindle accuracy. It's leveled and aligned as well as can be managed on a 3" concrete floor. Fortunately its construction is such that alignment is unaffected by out of level - it sits on three points. Even after 40 years (new April 1971) it passes all the turning tests. It HAS ben babied.

I went out and ran all the tests we've been discussing - RDM, TCM, test bar - over the last few days. Whatta PITA. I have a #4 Morse test bar that's very straight as tested between centers. The test bar in the spindle taper runs out 0.0003 on the end and sags (it's solid) about 0.0004" in its extended 12." According to the test bar when mounted between centers my tailstock is in 0.0006" lateral alignmentaway from the headstock. Yada Yada Yada.

Anyway I took a zillion readings, got a crick in my back, and scribbled pages of readings and sketches. It tells me my lathe is in pretty good shape. I can tidy up the pages of readings and stuff them in the lathe folder but what did I prove? Nothing really except the the RDM test was a little earier to manage than the first few times I ran it.

The RDM went OK but I had a hard time getting consistant repeat zero readings any closer than 0.0003". I used a 14" length of 1 1/4" cold rolled bar in the three jaw chuck. Naturally it ran out but the test instructions I used declared the run-out can be averaged out to compensate for any bar used for the test so long as it was round - hence, the cold roll. True, averagng works, but the accracy I attained suggests that 0.0015" alignment determination in 10" is about the limit. If that's good, fine, but not for me nor Mr Schlessinger.

The other tests repeated well and gave me intelligable readings. If the readings I took on my lathe required me to make an adjustment there is no doubt in my mind I could and prove it by subsequent tests.

So I conclude: pick the test that suits you and run it responsibly making sure to return to repeat zeroes for every reading you take. They all work but their success as with anything from boiling water to brain surgery is dependent on operator skill and diligence.

Yes I can take reliable readings to 0.0001". I have an elderly Federal 832 Gage amp and the EAS 1286 lever gage head designed for it. I can read to 0.000010" if need be.

J Tiers
06-01-2012, 09:13 AM
These tenths chasing theoretical evolutions have a way of meeting blunt reality in the form of older machine tools in different stages of decrepitude.

Yep.... a point I have made also.....

And a point AGAINST "one size cures all" types of "testing and alignment" methods such as RDM. Especially when they attempt to measure many things with a single measurement. (That's valid if you only want a "net result" measurement, but you cannot assess the effect of each, and may not be able to hit the minimum point you seek)

What's the point of a precision test with a machine that has other errors of 10x the precision of your measurement? Anything you find will be "lost in the noise".

Unless you have a machine that justifies the degree of precision that you are attempting to "survey into it", you are discussing angels on pinheads, and you are the pinhead...... At that point the exercise truly DOES become academic.

the errors may be due to wear, OR they may be due to the fact that your machine is really good only to a lower degree of accuracy...... the "1 thou machines" most here will have.

dian
06-01-2012, 09:28 AM
i still propose to look at it in pragmatic way:

so you decided to make something that needs a high degree of accuracy. you turn a little oversize, make a finishing cut and measure. you find one end is a little smaller than the other, by 0.005 mm lets say. so what do you do? take the work out of the chuck and start rollie-daddying your lathe?

not me, i put a shim under it (i might set up an indicator to check what im doing) and make another pass. its easy, i can get under the lathe with a car jack. it takes only a few minutes to have a cylinder as true as i can measure.

(you might want to adjust the top slide alternatively.)

J Tiers
06-01-2012, 10:34 PM
Perhaps the RTMCMTALATB test should be applied FIRST?

Then the machine can be corrected to the best extent possible and the cut will come out as right as that machine will ever make it....

or a length of abrasive paper can be pressed into service to remove slightly more from one end than the other..... perhaps by the approved "nutcracker" style device.

or, you can add up the thickness of hairs from Rollie, Gadgetbuilder, Forrest, myself, and whoever else was in or mentioned in this mud wallow, and take an average and range..... From that you can derive your maximum error resulting..... or you could just use them as shims......

Are we there yet?

dp
06-01-2012, 11:10 PM
In the case of RDM it would be useful to state some things at the outset:

The horizontal method assumes any discovered misalignment is the result of an otherwise aligned lathe having a twisted bed. If in fact the head is misaligned on the bed you will need other methods to correct the problem.

The vertical method assumes your head is properly aligned to the bed but the bed is curved vertically along its length, and that the bed is flexible enough that shimming one end is enough applied torque to remove that curve.

The horizontal method cannot distinguish between a twisted bed or unbalanced wear in the ways when applied to worn lathes.

The vertical method assumes facts not in evidence (bed flexibility). It also cannot distinguish a new curved bed from worn ways.

The RDM horizontal method is adequate for setting up a new lathe if one does not have a precision level, and even then, one cannot assume the lathe bed will be level when finished.

The RDM vertical method is probably not adequate to determine anything helpful as it depends on lathe inadequacy (flexible bed) to effect any alignment.

06-02-2012, 12:24 AM
I have removed this post because it was clearly over the line and I'm very sorry I made it.

Forrest's clear comparison of the various alignment methods at post 35 said all that was necessary concerning the issues raised in the prior thread. Unfortunately, I fixated on comparing TCT to RDM and became frustrated by my inability to understand the details.

This was a very interesting thread and I sincerely apologize for derailing it.

John

oldtiffie
06-02-2012, 02:01 AM
Hmmmmmmmmm.

Seems that this thread is on the down-hill run to oblivion as well.

J Tiers
06-02-2012, 03:14 AM
So it would seem......

And I am beginning to get rather bored with writing explanations of my point, that go in the blue......

I've seen RDM writeups for many many years. I had not looked for them for probably 8 or 10 years, and it seems that many of them, and the websites they were on, have gone away.. The Wasserman writup seems to have swept the rest away, almost everything anymore is a link to that.

nevertheless, a few references:

The classic Wasserman wreitup of the method complete with sheets of paper as shims..... obviously using RDM as a complete alignment system. It is, with the addition of paper shims, more-or-less the stock RDM system.

This guy has really no clue about what he calls the "dumbell method", he thinks TCT is done by putting the bar between centers...... I am assuming we all know that is not remotely the idea?
Much of the rest of his stuff seems quite sensible when read over quickly.... particularly the problems with indicator height while using RDM, shown farther down the page..... as the imperfect bar orbits, the indicator is sometimes on the proper spot, and sometimes above or below it, with attendant error.
He proposes a sweep tool which seems to carry a good bit of sag error and therefore complication to correct for it.
He is certainly not a starry-eyed proponent of RDM.

One of the responders here has good points

http://dir.groups.yahoo.com/group/atlas_craftsman/message/11774
Jon Elson has good comments in this.

As a rule, the adverse comments agree that the measurement method works, or at least can work. This is sensible, because it should work, although several point out that the roundness of the bar is an issue, and keeping the indicator on the max diameter is also.

Quite a few of the other googled references to RDM are from other boards and groups, where someone is trying to do it with long long rods.... at least one was a 15" long rod, which ought to be rather "interesting". I figure he got that from "somewhere" but I may not have found it yet.

I've said what I have said, and unless I think of something else, or find more references, I will probably let this lie down and die.

06-02-2012, 12:13 PM
So it would seem......

This guy has really no clue about what he calls the "dumbell method", he thinks TCT is done by putting the bar between centers...... I am assuming we all know that is not remotely the idea?
Much of the rest of his stuff seems quite sensible when read over quickly.... particularly the problems with indicator height while using RDM, shown farther down the page..... as the imperfect bar orbits, the indicator is sometimes on the proper spot, and sometimes above or below it, with attendant error.
He proposes a sweep tool which seems to carry a good bit of sag error and therefore complication to correct for it.
He is certainly not a starry-eyed proponent of RDM.

...
As a rule, the adverse comments agree that the measurement method works, or at least can work. This is sensible, because it should work, although several point out that the roundness of the bar is an issue, and keeping the indicator on the max diameter is also.

Jerry,

You have identified two clear issues with the RDM concept: cosine error due to bar runout and bar roundness.

I believe I minimized cosine error during setup (see my RDM description) so I think I dodged the indicator height issue. My 3 jaw chucks have about a thou runout (if not a 4 jaw will) and the procedure given reduces the runout at the far end to a thou or so. This simulates the accurate test bar I wish I could afford. Real test bar runout is spec'ed at 4/10 so a thou or so is not too different. Cosine error is small for this level of runout and I contend it can be ignored.

Roundness is a more interesting problem. It looks like the lathe makes items round. But roundness is relative and a surprisingly difficult property to measure when you want something seriously round. Eccentricity of turned work depends on the history (heating, bending, etc.) of the item as well as the machine it is turned on.

I can't measure roundness accurately in my shop but I believe I can make things round enough so the out of roundness doesn't appreciably affect the RDM readings. Rollie (of RDM fame) made a presentation on lapping to our local machinist club, similar to his video on the NEMES site. What he said made sense: lapping is a statistical process in which the lap and the work get rounder and in the limit the outcome would be a perfect circle.

You can actually feel the eccentricity of items turned on the lathe as lapping begins - the lap twists in your hand with each rotation. This eccentricity goes away pretty quickly because much of it initially is tooling marks rather than solid metal. In addition, running the lap axially causes the diameter to become constant over the lapped area, one can feel differences of less than 1/10 so it's easy to convert a bar that looks round and constant diameter to one that IS, within better limits than I can measure. As a fringe benefit, the surface noise as seen by the indicator is greatly reduced.

The point is, we're stuck with the indicator error but we can minimize noise seen by the indicator due to eccentricity, differing diameters along the bar, and surface noise. And this can be done cheaply and fairly quickly in the home shop. Lapping should be done on ground bars as well as those turned in the lathe -- grinding can introduce eccentricity errors too, depending on the process. The improvement isn't easily quantified, it is simply a best effort to minimize known sources of noise/error in the reading.

Having done the easy and obvious things to reduce measurement noise, we come to bar length. The test bar is like a noiseless amplifier so the longer it is the better -- the signal gets larger but noise from the sources above is constant. Unfortunately, bar length brings up the sag issue where one must calculate the sag, so any error in the sag calculation affects the vertical result directly. Mother Nature rules here so a large diameter tube would be the best choice.

On the outside chance you haven't totally lost interest, perhaps you'd like to discuss how you handle eccentricity of your turned collars, surface noise, shorter bar length and how the combination affects the results of TCT.

J Tiers
06-02-2012, 12:39 PM
On the outside chance you haven't totally lost interest, perhaps you'd like to discuss how you handle eccentricity of your turned collars, surface noise, shorter bar length and how the combination affects the results of TCT.

Well, my approach is that eccentricity of turned surfaces is part and parcel of the machine..... it's the noise floor. Don't try to polish a turd.

Finish is as good as you can get it, and with appropriate material and tooling can be quite good... I don't see that as a problem, but if you do, there is nothing saying you can't use a TPG instead of a cutter......

Bar length is what it is, and longer will improve resolution in the results with ANY bar-type method, if you can handle any stiffness-related issues so that they are not an issue. That may be problematic......