Thread: Was this just good luck?

I had a huge amount of trouble setting up my AL320G lathe, I don’t have a precision level but I tried every method I could think of or find, but the best I could get it was turning about .002in taper over 100mm. sorry about the mixed measurements, my micrometer is imperial and the lathe dimensions are metric.

Lying in bed the other morning I got to thinking how I might fix the problem, so much so I got up at 5am and headed for the shed, and believe it or not I got it sorted before breakfast.

This is what I did. The distance between the lathe fixing points is 722mm, so I mounted a piece of 1in. bar in the lathe and turned a 73mm section, the taper was .0015in. being larger at the tailstock end. My theory was, if I shimmed the front mounting pad at the tailstock end by 10 times the taper divided by 2 (.0075in)it might fix the problem. The nearest shim steel I had was .008in so I used that.
When I turned the 73mm section again, no taper so I doubled it to 150mm still no taper exactly the same measurement each end, after all my previous mucking around I was over the moon that I could fix it so easily.
While I was on a roll I thought I would check to see how far out the tailstock was, so I turned a 400mm between the 3 jaw chuck and the tailstock centre, took a measurement each end and they were exactly the same, infact there was no variation anywhere along the length, I thought it might have flexed in the middle.

I think I just fluked this, because when I think of it my original theory doesn’t take into account the depth of the mounting points from front to back, or the height from the base to the bed, and as I see it these two things would have a big influence on the result.

Any way my lathe is as nearer to perfect than I ever hoped to get it so I’m as happy as a pig in the proverbial.

You may have fixed the problem in a sense, but the problem may only be solved for one particular position of the tail stock...

Assuming the headstock is correct (and it should be as a new machine), then the other two variables for turning a taper are the bed and the tailstock alignment. If the bed is twisted, the twist can be countered by moving the tailstock across until the taper is gone. This is great until you move the tailstock inwards or outwards, where the amount of twist is different. Likewise, if the tailstock is wrong, you can fix it by twisting the bed until the taper is gone again. But this will only work for the tailstock where it is, as soon as you move it away from that spot, the twist is different and the adjustment will not be correct.

Thus, the only way to get it right is to make sure the bed is flat via another means, ie, a level. Once the bed is proven to be correct, you then adjust the tailstock to remove any tapers.

The way to check whether your setup is (by chance) actually correct because of this shimming is to cut some test rods with different tailstock locations. if the lathe refuses to turn a taper, no matter where the tailstock is, then both the bed and the tailstock are correct.

I see where you are coming from, But I thought that because I could turn a 150mm section that was unsupported at the tailstock end and have no taper that the headstock and the bed must be right (or close to). After that I checked the tailstock by turning a piece much longer that was supported at the tailstock end. i still had no taper so made no attempt to adjust the tailstock.
I will try few pieces of different lengths between centres to see what results I get.

A lot of things would flex when cutting unsupported 150mm from the chuck. Using a dial guage would be better, but you need to depend upon the work piece (or measuring piece as it were) being mounted perfectly co-axially with the spindle in order to be reliable, so if this in ye olde trusty 3 jaw chuck, you can probably disregard those measurement techniques. Really the best method is the tried and true - level the bed and set the tailstock separately. Even getting the headstock right should really be done with a bed that is level, its the basis for all other components after all.

> Assuming the headstock is correct (and it should be as a new machine), then the other two variables for turning a taper are the bed and the tailstock alignment.

wrong

> If the bed is twisted, the twist can be countered by moving the tailstock across until the taper is gone. This is great until you move the tailstock inwards or outwards, where the amount of twist is different. Likewise, if the tailstock is wrong, you can fix it by twisting the bed until the taper is gone again. But this will only work for the tailstock where it is, as soon as you move it away from that spot, the twist is different and the adjustment will not be correct.

wrong

This alignment is done without tailstock support. It must under no circumstances be done with tailstock support!

To untwist the lathe bed (if you are SURE the headstock to bed alignment is ok) you must hold a spool shaped workpiece in the headstock. Put the tailstock in the drawer. Measure for taper between both spool ends, and correct by shimming at the right side of bed.

Only once this basic alignment is done, can you proceed to adjust the Tailstock. If you try the other way around, you will never end up with a properly aligned lathe. Chris

PS: hint. If you follow the order of adjustments given in a typical inspection record, you will be fine.

Care to explain a little more than just wrong...?

As I said in my post, cutting a test piece (your spool shaped example) over a suitable distance will lead to far too much flex in the test piece and give an incorrect result. If you shorten it up to make it rigid, the measurements will be pointless as youre not sampling over a long enough section of bed. At the very least, for your example you should be using a dial indicator, not actually cutting the test piece to eliminate those forces.

Originally Posted by scottyd

Care to explain a little more than just wrong...?

I shall try, but it is not as clearly black and white as one would like...

> As I said in my post, cutting a test piece (your spool shaped example) over a suitable distance will lead to far too much flex in the test piece and give an incorrect result.

The test piece is always fat and short to minimise sag and deflection. For a bench lathe, something like more than 40mm thick and protruding 100mm from the chuck will perfectly do. You must use a very sharp tool: either a freshly ground and carefully honed HSS, or a polished Alu type insert like DCGT or CCGT or the like witch have extremely positive rake and are razor sharp. The sharper the tool, the smaller the chip that you can cut with it in a controlled manner, the lower the cutting forces, the less the deflection of workpiece and the overall flex of the machine.


> If you shorten it up to make it rigid, the measurements will be pointless as youre not sampling over a long enough section of bed.

It does not matter. Any overall twist in the bed will be equally and uniformely present also in a short section of the bed. If your lathe seems to have differing rates of twist along the bed, then you are really not observing twist; but most likely a defect like wear or warping of the bed casting (not seasoned before grinding to save money during manufacturing). Wear or warping cannot be corrected by any adjustments, because such defects are not uniform over the length of the bed - if wear gets too bad you either have to learn to live with it, or overhaul the lathe, or use the particular lathe for tasks that do not need the high degree of accuracy, or replace the lathe.


> At the very least, for your example you should be using a dial indicator, not actually cutting the test piece to eliminate those forces.

See, the dial indicator is what you use for untwisting the bed, which is often simply called "lathe alignment" because it is usually the only alignment that is required when installing a new lathe. The test cut is actually done to confirm a successful lathe alignment, and if necessary to do some tiny fine final adjustment to it. Since the final purpose of any lathe alignment is unquestionably for the lathe cut parallel, and not to simply enjoy looking at nice low tolerance figures on an inspection record. Notice that such a test cut is a requirement of most inspection records, and for good reason it is usually the last step in such record.

But maybe we should differentiate between the case of a near new lathe, and the case of a worn out or warped lathe.

On a near new lathe, it makes sense to very finely correct bed twist for the slight cutting forces used to turn the spool, because these are exactly the cutting forces that you will later apply when applying the finishing cut to a shaft or a bushing that you expect to be dead parallel.

On a worn lathe, after aligning the lathe the test cut is likely to show a major taper. This is mainly because of irregular wear along the bed ways. At first thought, it may seem to makes sense to "untwist" the bed according to the test cut close to the chuck, because this is where later you will do 90% of all your workpieces, since the vast majority of workpieces are usually short. But doing so would mean that any longer worpieces would invariably end up with a very bad taper. And vise versa. In truth, there is simply nothing you can do by way of aligning. An experienced machinist may still be able to produce reasonably parallel workpieces on such worn machine, but it takes considerable experience and skill to "correct on the fly" for wear. In the old days, this skill was what did set a good machinist apart. Chris

Originally Posted by cba_melbourne

I shall try, but it is not as clearly black and white as one would like...

> As I said in my post, cutting a test piece (your spool shaped example) over a suitable distance will lead to far too much flex in the test piece and give an incorrect result.

The test piece is always fat and short to minimise sag and deflection. For a bench lathe, something like more than 40mm thick and protruding 100mm from the chuck will perfectly do. You must use a very sharp tool: either a freshly ground and carefully honed HSS, or a polished Alu type insert like DCGT or CCGT or the like witch have extremely positive rake and are razor sharp. The sharper the tool, the smaller the chip that you can cut with it in a controlled manner, the lower the cutting forces, the less the deflection of workpiece and the overall flex of the machine.


> If you shorten it up to make it rigid, the measurements will be pointless as youre not sampling over a long enough section of bed.

It does not matter. Any overall twist in the bed will be equally and uniformely present also in a short section of the bed. If your lathe seems to have differing rates of twist along the bed, then you are really not observing twist; but most likely a defect like wear or warping of the bed casting (not seasoned before grinding to save money during manufacturing). Wear or warping cannot be corrected by any adjustments, because such defects are not uniform over the length of the bed - if wear gets too bad you either have to learn to live with it, or overhaul the lathe, or use the particular lathe for tasks that do not need the high degree of accuracy, or replace the lathe.

I think here is where our opinions are differing. When I say that the twist varies based on where the tailstock is, it is because the amount of twist being exhibited by the bed is determined by how much bed is between the head and tailstocks. Twist is measured in degrees/meter or mm/m. Lets say there is a 10mm/m twist in the bed (an extreme example) and the bed is 1m long between centres. If you cut a test piece that is 500mm long, then the test piece will be cut / dialed to be 5mm out. if you cut a test piece as per your example, then the evident error will be just 1mm. Obviously, these are large figures, but a loss of resolution of 90% compared to what the bed is actually twisted to is nowhere near good enough when people are chasing .02mm/m leveling or better with the bed via other means. To achieve the same accuracy using your suggested technique would mean:
- you will need to measure and repeat the test cuts at better than 0.002mm over the 100mm test piece.
- the test piece needs to deflect significantly less than 0.002mm. Any deflection will yeild 10x that error in setting the bed.

These issues are amplified in a longer bed.

Originally Posted by cba_melbourne

> At the very least, for your example you should be using a dial indicator, not actually cutting the test piece to eliminate those forces.

See, the dial indicator is what you use for untwisting the bed, which is often simply called "lathe alignment" because it is usually the only alignment that is required when installing a new lathe. The test cut is actually done to confirm a successful lathe alignment, and if necessary to do some tiny fine final adjustment to it. Since the final purpose of any lathe alignment is unquestionably for the lathe cut parallel, and not to simply enjoy looking at nice low tolerance figures on an inspection record. Notice that such a test cut is a requirement of most inspection records, and for good reason it is usually the last step in such record.

But maybe we should differentiate between the case of a near new lathe, and the case of a worn out or warped lathe.

On a near new lathe, it makes sense to very finely correct bed twist for the slight cutting forces used to turn the spool, because these are exactly the cutting forces that you will later apply when applying the finishing cut to a shaft or a bushing that you expect to be dead parallel.

On a worn lathe, after aligning the lathe the test cut is likely to show a major taper. This is mainly because of irregular wear along the bed ways. At first thought, it may seem to makes sense to "untwist" the bed according to the test cut close to the chuck, because this is where later you will do 90% of all your workpieces, since the vast majority of workpieces are usually short. But doing so would mean that any longer worpieces would invariably end up with a very bad taper. And vise versa. In truth, there is simply nothing you can do by way of aligning. An experienced machinist may still be able to produce reasonably parallel workpieces on such worn machine, but it takes considerable experience and skill to "correct on the fly" for wear. In the old days, this skill was what did set a good machinist apart. Chris

You are right in seeking a parallel cut as the end goal, but if you are turning a long piece unsupported, it can easily end up tapered irrespective of the alignment of the lathe. Final test cuts are always taken between centres for this reason. For any cutting task requiring precision, the workpiece should be as well supported as possible. I disagree about countering cutting forces with a slightly misaligned bed because of this reason. The best way to get repeatability is to have the machine as aligned as possible. As I illustrated above, if you deliberately misalign the bed, the effect of that error varies with where the cut / tailstock is on the bed, as well as variations with materials and cutting force generated.

> using your suggested technique would mean:
> - you will need to measure and repeat the test cuts at better than 0.002mm over the 100mm test piece.
> - the test piece needs to deflect significantly less than 0.002mm. Any deflection will yeild 10x that error in setting the bed.

It is not my suggested technique, it is the proven and universally accepted technique.

And you cannot align a lathe to anything near 0.002mm as you seem to hope for. You can not even achieve such a degree of roundness, not even with the most expensive precision class spindle bearings found in a toolroom precision class Schaublin and Hardinge. Lathe alignment is not an end in itself. The degree of accuracy by which you aim to align your lathe must bear some relation to reality. Most inspection records will call for something like less than 0.03mm taper over 100mm. That is something achievable, something that you can easily measure with an ordinary micrometer screw, without having to worry about the heat radiation from your or your hand or your warm breath falsifying the reading.

A little experiment to open your eyes and get the above 0.03mm over 100mm into context: make a test cut and measure it. Then take another test cut, but this time during the cut press with your little finger as hard as you can against the side of the tailstock to simulste the bed twist caused by a force of one or two kilos. Now measure again. On a hobby class bench lathe, unless you have a heavy industrial lathe, you now have up to 0.05mm extra taper. That is how flexible your lathe is. Now you understand why any aligning better than 0.01mm over 100mm is just a momentary theoretic exercise, with no relation to practical work.


> These issues are amplified in a longer bed.

They are not. A taper of 0.1mm/m remains the same, regardless how long the bed is.


> You are right in seeking a parallel cut as the end goal, but if you are turning a long piece unsupported, it can easily end up tapered irrespective of the alignment of the lathe. Final test cuts are always taken between centres for this reason.

Final test cuts in the context of bed twist alignment are NEVER and must NEVER be taken between centers. It would completely falsify the result. The tailstok is only aligned AFTER everything else is aligned. I highly recommend you get hold of a copy of "Testing Machine Tools" by Georg Schlesinger, great reading and the ultimate reference in matters of alignment. Or at least, stick to the order in which the individual tests are listed in the inspection record.


> For any cutting task requiring precision, the workpiece should be as well supported as possible.

Correct. But aligning a lathe is not a simple cutting task, is it?


> I disagree about countering cutting forces with a slightly misaligned bed because of this reason. The best way to get repeatability is to have the machine as aligned as possible. As I illustrated above, if you deliberately misalign the bed, the effect of that error varies with where the cut / tailstock is on the bed, as well as variations with materials and cutting force generated.[/QUOTE]

I have never suggested to misalign the bed, far from it. I have suggested to do the fnal fine adjustment of the bed based on a test cut. And only provided the lathe is in reasonably good nick and not already worn out. If you are into accurate work that is what you do anyway from time to time, to correct for seasonal changes caused by things like floor movement, room humidity, average room temperature etc. You take a cut on your spool, and correct with the bed levelling screws at the tailstock end.

As for the long workpieces you seem mainly concerned with: the reality is that longish workpieces are turned supported at both ends (a differnet setup geometry to the one used for bed alignment). If you need it to be dead accurate at both ends you either correct taper with the sideways adjustment of the tailstock, or turn the workpiece end to end for the finishing cut. Of course you still may turn a larger diameter at the center of your workpiece beacause of combined gravity sag and deflection from cutting force, and in a long slender workpiece this is not unlikely to be a greater error than any taper error at the ends. Chris

It seems to me that the 3 jaw chuck makes no difference, because once you have turned the workpiece and not removed it from the chuck, it would have to give a true indication of the lathe accuracy, the problem arrises if you remove the job from the chuck and try to rechuck it.

I have done many test cuts of various lengths, both supported and unsupportd at the tailstock, in the chuck and between centres and I am satisified that I have less than .0005in taper over 350mm and any taper at 100mm length I am unable to measure with my instruments.
I am not going to mess with any further adjustments because I'll probably never get it that good again.

To do these test cuts I used 25mm steel rod, very sharp HSS cutting tool and turned off .001in for my final cut, deflection did not seem to be an issue as even on the longest piece the measurement was the same for the whole length.

what you achieved by cutting a piece,working out amount of taper and distance between mounting is virtually what you would do with a engineers level.

You would place it on the bed or suitable area and put shims/feelers under it until the bubble was centred,divide the length of your level into the mounting distance and then multiply amount of shim/feeler by this.This will then give you a starting point for packing under the feet,the process needs to be repeated until your happy with the result.

If your happy with the method you used and it working that's all that matters.