Thread cutting tolerances ?

[NealebParticipant @nealeb]

One small (possibly very small) point is that the approximate threads produced via 63T or 21T gears are fine for nuts and bolts or similar fixings. However, in this case, the thread is being used to provide a linear drive, albeit to the periphery of a gear, and it might be possible that with many rotations there will be an accumulated error. If the gear only rotates one turn or less and just back and forth, probably not an issue. Just a thought.

[Nealeb]

On Nealeb Said:

However, in this case, the thread is being used to provide a linear drive

Please could you provide a numerical value for the pitch error produced by the two approximations you quote.

Please could you explain the ‘linear drive’ part of your post. It is a worm and wheel that he is making.

With a (single start) worm and wheel, the worm goes round one turn and the wheel advances one tooth. If the wheel has 40 teeth, once the worm has rotated 40 turns the wheel is back to the beginning. Please could you explain how any pitch error produced by using approximate gearing will disrupt this process.

[duncan webster 1Participant @duncanwebster1]

Removed

[DC31k]

On DC31k Said:

[…] With a (single start) worm and wheel, the worm goes round one turn and the wheel advances one tooth. If the wheel has 40 teeth, once the worm has rotated 40 turns the wheel is back to the beginning. Please could you explain how any pitch error produced by using approximate gearing will disrupt this process.

It does, of course, rather depend upon how closely one is looking.

‘Disruption’ is reasonably well discussed here:

https://astrojolo.com/gears/mount-periodic-error/

MichaelG.

[DC31k]

On DC31k Said:

On Nealeb Said:

However, in this case, the thread is being used to provide a linear drive

Please could you provide a numerical value for the pitch error produced by the two approximations you quote.

Please could you explain the ‘linear drive’ part of your post. It is a worm and wheel that he is making.

With a (single start) worm and wheel, the worm goes round one turn and the wheel advances one tooth. If the wheel has 40 teeth, once the worm has rotated 40 turns the wheel is back to the beginning. Please could you explain how any pitch error produced by using approximate gearing will disrupt this process.

An exact metric-imperial conversion for a screw-cutting gear train, as already mentioned, needs a 127T gear somewhere. A 63T is used as an approximation – often an acceptable approximation – although ideally should be 63.5T. Clearly not possible. The error is in the ratio 63:63.5, about 0.78%. This is widely accepted for short threads, but it’s not exact. Numerically, the same percentage error exists for 21T. There may well be better approximate gear trains but these generally involve gears which are not in the standard set. In the current case, if our “approximate” worm is used to rotate a 63t wheel, after 63 worm rotations the wheel rotation will be about a half-tooth out (either more or less, depending on exactly which way the error has been incorporated).

Yes, it is a worm and wheel setup. However, I suggest that it is incorrect to think “one turn, one tooth”. When the worm pitch is exactly correct, this is true, but we do not have an exact pitch. The worm pitch is very close to but not the same as the tooth spacing. I also agree that overall, this is a rotating and not linear drive, but at the point of contact it is a linear drive, or at least a very good approximation to one. If we take this “approximate” worm and make a nut to fit, then that nut will advance by one thread pitch for every turn of the worm. Now cut away that nut so that only a small slice containing a few threads exists – so it looks like a thin plate with teeth cut in the edge – and it will still move one thread pitch per turn. That is pretty much what we have with the worm wheel – the drive is linear in the region of the tooth engagement although the effect overall is to rotate the wheel. At the same time, we modify the wheel’s tooth form from rack (in the case of a nut) to involute to give better match between worm and wheel. Although, as I say, we do not have an exact tooth form anyway as the worm pitch and wheel tooth pitch do not match exactly. Nevertheless, I’m sure the worm pitch error will not stop the whole system working in a purely mechanical sense, but rotation of the wheel will not be exactly correct. If the wheel has to move a precise amount, or if it always rotates in the same direction rather than partial rotations back and forth, then the error may be significant. If the goal is not exact rotation, or if there is some kind of rotary encoder to provide feedback, this may not be important.

[Brian WoodParticipant @brianwood45127]

I tested out this aspect some years ago to cut three worms of varying pitch accuracy on the same shaft, engaged with ‘worm’ wheels of identical tooth count and ran the combination to see if any discrepancy became apparent in the wheel displacement after many revolutions, having started them off with marked teeth. They were all free to run as dictated by the worms

The worms were cut with gearing that ranged from grossly crude to the best I could achieve short of using a leadscrew wheel of 127 teeth, which I don’t have.

The result rather surprised me—there was no displacement variation after a long period of running of about 15 minutes or so. I then reversed the wheels on their axles and repeated the test, with the same result.

The worm pitch itself made no difference to the outcome. I concluded then that the quality of the tooth count was the governing factor in the outcome. It seem that in practice one turn one tooth does apply.

Brian

 

[duncan webster 1Participant @duncanwebster1]

I think the 63t gear should be used as a driver. It is not that it is approximately half of 127.

I initially thought pitch error would give rixe to error in rotation, but now I don’t, hence my removed post above. What I reckon now is that at some point as different wheel teeth engage with the worm, the wheel will jump back into synch.

 

[HopperParticipant @hopper]

One pitch is one pitch, regardless of whether it is exactly the correct pitch or not.

So in exactly one revolution of the worm, the worm thread returns to the exact same position it started from.

So if the worm thread is in the exact same position it started from, the point of contact between the worm thread and the worm wheel’s next tooth must be exactly the same. And so on for Brian’s 15 minutes of revolutions, ad infinitum.

The consistency of pitch of the teeth on the wormwheel would be of much greater importance. If they were inconsistently spaced, rotation of the wormwheel would be proportionately inconsistent. Which in practical terms would be a negligible amount in most cases.

 

[duncan webster 1]

On duncan webster 1 Said:

I think the 63t gear should be used as a driver. It is not that it is approximately half of 127.

 

I think (or rather Google says) the 63 tooth is used in a 80/63 combo that gives ratio of 1.2698, an error of  0.0125 per cent from the perfect 1.27 ratio of a 127 tooth gear. I reckon I could live with that.

ISTR Neil did an article in MEW in recent years on the 63T gear but don’t remember what it said, as I use Martin Cleeve’s chart for using the original Myford change gears compounded for metric threads so no need for 63 or 21 or 127 teeth exotica. Cleeve’s charts’ error rates are 1 in 3,000 to 1 in 8,000. Again plenty close enough for me.

I would expect that the 80 in the 80/63 could be achieved by compound gearing a 40 to run at double speed etc.

[Brian WoodParticipant @brianwood45127]

Hopper,

I believe it was a Southbend approximation that they offered in place of the 100/127 compound wheel set on a ball bearing trunnion that the lathe had space for, but I imagine it was costly to buy as a spare part. As you say, it was a good one.

Another of different values that Hardinge used was set as 37/47. The error was slightly more that the Southbend offering but it had the great value of reduced size and could be easily accommodated on smaller sized lathes

Brian

[Charles LamontParticipant @charleslamont71117]

Really close approximations can be achieved by using 21 and 63 tooth together, both as driven gears. Typical errors of 1 part in 10,000 are possible. For an 8 TPI leadscrew, eg:  meshing: driver > driven.   stud: driven / driver. (idler).

0.999937 mm

25 > 21 / 20 > 63 / 50 > 60

2.999811 mm:

25 > 63 / 50 > (45) > 21

Both calculated by the ML7gears program I mentioned earlier.

[Niels GloverParticipant @nielsglover18824]

Gents,

I’ve arrived here because I’ve recently been clearing out some of my old dusty boxes with a view to ‘rationalising’ the stuff I have. That led me to wondering what the current thinking is on the age-old ‘cutting metric threads on an imperial lathe’ question.

l’ve been in the electro-mechanical fix-it business during my working life but have never worked in a shop or trained as a machinist in any way. Thus, when my Uncle’s Atlas 10F was passed to me some 35 years ago I was starting from a true ‘amateur’ status. To think, he had this machine for 30 years so it’s now been in the family for 65 years – astonishing how time flies.

Way back then, when the whole idea of metrication was much newer than today, and with, more or less, only imperial machines available for the amateur, ‘metric threading’ all seemed a bit of a black art to me. I had mastered ‘ordinary’ threading (gear trains etc.) but I remember the terminology of the time to be confusing in often referring to ‘conversion’ gears or ‘converting’ the lathe to cut metric threads. This was, until I spoke to a retired neighbour of ours who was an ex-Molins engineer.

He said to me “what you need to understand is that you can’t cut a metric thread on an imperial lathe”.

Of course I’m “yes, but, yes, but….”

He said “what you can do is cut a non-standard imperial thread of 25.4 tpi”

With that one golden nugget of wisdom, everything fell into place as all the standard calculations still apply. He and I became good pals for more than 30 years. Sadly no longer with us.

I have been interested to read the various opinions on the forum regarding ‘approximate’ threading using standard gears. Those seem to vary from not particularly favourable to the other end of the scale suggesting that a 127 combination isn’t even needed. I wondered why that difference might be. Could it be down to which machine you have. I.e. the combination of lead screw pitch and the standard gear-set supplied? I have to say that I never found the approximate ‘metric’ threading, using the Atlas recommendations to be satisfactory. Yes, it sort of worked, but either a bit loose, or tight depending upon the approximations.

I thought salvation was at hand. When my Uncle passed away my Aunt asked me to pick up a couple of boxes of Uncle’s ‘stuff’ to sort through and keep or dispose of. Lo and behold! In the bottom of a box was a 127/50 gear pair in what I’m sure is 20DP. They had obviously been adapted for the Atlas. Problem solved! …..Until I tried to use them. Even at 20DP they were far too big to mount on the Atlas banjo and have any room left in the gear case to do anything with them. At that moment I realised two things. Firstly, my Uncle had probably gone through the same exercise in getting these gears, adapting them to the Atlas and then realising they were impractical – hence ending up in the bottom of the box and secondly, it became clear why Atlas didn’t offer a 127 gear for the 10F. In the standard Atlas 16DP it would be approaching the size of a dinner plate!

All was not lost, however, because eventually I was able to use these gears as indexers to make myself a new 127/50 pair in 26DP which would fit. I therefore use this pair to run together as a ‘fixed ratio’ and use the standard Atlas gears to multiply or divide the fixed ratio to give me the pitches I want. I can’t say this would give me any pitch but I haven’t yet found anything I needed which I couldn’t do.

If anybody has bothered to read this far, they might be wondering what this ramble has to do with clearing out the workshop. The thing is that I have, again, come across this original 127/50 gear pair. I had actually put them on the ‘next trip to the dump’ pile. Sitting with a thoughtful cuppa, it occurred to me that they might be useful to someone?
Anyway, that’s the ultimate point. If anyone can use these gears I would be delighted to post them FOC to the first raised hand. If I get no takers, that’s OK too because it would ease my conscience considerably over throwing them out. As said, they are adapted for the Atlas, but I’m sure a happy hour’s project time would see them sleeved to fit anything.

As a final aside, another piece of wisdom I wish I’d had years ago is what to keep and what not to. I have stuff I’ve kept for years, never used and never will use. Why do we do that? I can only think that 30 years ago I didn’t know if I needed something. It’s the wisdom of years which allows you to realise you didn’t need whatever it was and never will. Even so, it’s still hard to fathom. Two extreme examples in my case. One is a dinky little Bunerd 3 in 3 jaw chuck. Nice little thing. No perceptible wear. Probably had little use, if any, but a 3” chuck on the Atlas? I doubt it! At the other end of the scale is a No. 36 Jacobs drill chuck. This thing opens out to about 3/4”. A 3/4” straight shank drill in the Atlas? Sounds like a recipe for disaster to me! But still I’ve kept this stuff for decades??

Anyway, the time is now! Two more items for the chuckout pile. (Did anybody spot that one? The ‘chuck-out’ pile?…..Oh, please yourselves…)

[Brian WoodParticipant @brianwood45127]

Is there a page missing on this topic? Yesterday an interesting exchange, to me at least, had contributions from Neil Wyatt, Hopper and Pete Rimmer, as well as my own observations of those comments.

Today, all that has completely vanished–where have those posts gone?

Brian

[Brian Wood]

On Brian Wood Said:

Is there a page missing on this topic? Yesterday an interesting exchange, to me at least, had contributions from Neil Wyatt, Hopper and Pete Rimmer, as well as my own observations of those comments.

Today, all that has completely vanished–where have those posts gone?

Brian

That would have been the similar thread titled “Boxford aud metric gears” ,which meandered into general discussion of such things as 63T and  127/135T gears. Seems to be several threads recently on metric screwcutting on imperial lathes. Sign of the times I guess, as more drawings are provided in metric, and younger model engineers are not so keen on the king’s measure.

Brian, another similar thread popped up overnight, “Early Myford Super 7B metric conversion” that you may  be able to throw some light upon, relating to gearing for the Myford early QC gearbox vs late. https://www.model-engineer.co.uk/forums/topic/early-myford-super-7b-metric-conversion/

[Brian WoodParticipant @brianwood45127]

Thank you Hopper, more damage to the grey matter it seems just from being old too, rather unfair!

Brian

[HopperParticipant @hopper]

I know the feeling — very well!

[Michael GilliganParticipant @michaelgilligan61133]

Neils Glover

I’ve “read that far”

Please look-out for a personal message from me !

MichaelG.

[Neil WyattModerator @neilwyatt]

I hate to point this out now the job is done… but if you have a 3D printer why didn’t you just print the required 63-tooth gear? (Yes, this does work perfectly well).

Neil

[mark costello 1Participant @markcostello1]

If He does not have a printer, but has a DRO with bolt hole circle function a gear can be easily done.

[Martin of WickParticipant @martinofwick]

Having perused this thread, I am still none the wiser as to what is an acceptable tolerance for thread cutting (note acceptable as opposed to the minimum).

Say I want to cut a metric thread pitch =1 and my best change gear selection will produce said thread with an error of  +1%.

I take that to mean there will be 1.01 threads per mm. Is that acceptable?  Is it OK for a 6mm nut but a disaster for a 10mm locating hole?

Is there some rule of thumb to assess acceptable error for pitch or length?

Clearly the smaller error the better, but how small is small enough  for goodnuff?

[Martin of Wick]

On Martin of Wick Said:

[…]Is there some rule of thumb to assess acceptable error for pitch or length?

[…]

Clearly the smaller error the better, but how small is small enough  for goodnuff?

How long is a piece of string ??

If you are making non-critical ‘nuts & bolts’ then near-enough is good enough

But if you’re making a leads-screw or a micrometer, it’s rather different !

MichaelG.

.

Edit: __ and if it really, seriously, matters … this is approaching perfection:

https://lathes.co.uk/bryantsymons/

[Charles LamontParticipant @charleslamont71117]

OK. For the sake of argument let us take a screw and nut with a 1 mm pitch. The parts are made with the least clearance to allow assembly by hand, if the thread pitches match. However, the 10mm long nut has a 1% pitch error and the parts will not assemble by hand. How much extra clearance is needed to allow them to fit easily?

Well, we have to find 1% of 10mm, or 0.1mm of axial clearance. Now as it happens, despite being a metric pitch, our thread has a 55° Whitworth form, which is convenient for us because the tangent of 27.5° is pretty close to 1/2. If we increase the internal radius of the the nut by 0.1mm we will gain 0.05mm axial clearance at each end of the nut. So the nut thread diameter needs to be increased by twice the total axial pitch error over the length in question. In this case, that would represent about 16% of the thread depth.

In practice, an M6 has a 1mm pitch, and a standard M6 nut is 5mm thick and a clearance of 0.1mm on diameter feels to me about typical of a commercial nut and bolt. It they will fit together, then when properly tightened there will be enough plastic deformation of the threads to take up a 1% pitch error.

It would take a whole different order of analysis to consider how much interference would allow the nut to be fitted satisfactorily using a spanner, and how much would be likely to result in the threads galling and seizing.

[JasonBModerator @jasonb]

Charles how does altering the diameter work when it is the flanks that will be in contact?

This is why it is often said that a thread with slight pitch error will fit a nut which is relatively short but if you want to screw that same thread into a deep tapped hole it will bind

Quick stetch here, plenty of clearance between roots and crests ignoring any rounding over. Screw in the middle is 1mm pitch and the nut is 1.01mm pitch.

You can see that there is clearance on the right until the point marked in green so a short nut would be OK but anything longer will bind

If you are saying reduce thread dia but keep thread depth the same you will get the parts to fit for a longer length but introducing more slop to compensate for the wrong pitch means there is a lot of clearance and the actual point in contact is small (blue line) This is 6mm x 1mm pitch into what would be a 6.4mm x 1.01mm pitch hole

 

[Martin of WickParticipant @martinofwick]

So as a first shot to distil some acceptable general tolerance  rules, ignoring special cases and exceptional conditions …?

 

pitch error from zero to .0001% For motion control, common or garden use (as opposed to scientific precision)

pitch error between .0001% to .001% for fasteners with contact area up to 20 x pitch

pitch error between .001% to .01% for short form fastenings (ie, standard nuts and suchlike)

 

 

[Martin of Wick]

On Martin of Wick Said:

So as a first shot to distil some acceptable general tolerance  rules, ignoring special cases and exceptional conditions …?

 

pitch error from zero to .0001% For motion control, common or garden use (as opposed to scientific precision)

pitch error between .0001% to .001% for fasteners with contact area up to 20 x pitch

pitch error between .001% to .01% for short form fastenings (ie, standard nuts and suchlike)

 

 

Though I like the idea of a rule of thumb I’d have to think about it, maybe arguing that any pitch error is a bodge that should always be avoided!

Anyway, thinking on the hoof, I believe pitch error causes three side-effects, which might be serious, depending on the circumstances:

• A nut of ‘n’ threads with pitch error will jamb on a perfect male.   May not matter if the nut is less than ‘n’ threads long, because the error never causes a jamb – the two threads have clearance.
• However, the error means that a small part of the faulty thread will engage before the rest, such that the force normally taken by several turns concentrates on one.   This could cause progressive stripping, suggesting pitch error is less acceptable on weight bearing components, and not at all if the thread is safety critical.  Except progressive stripping is less likely if the nut and screw are ductile, and the pitch error is small, because the overloaded section of thread will move plastically allowing the rest to engage as well.   Is plastic adaptation to pitch error acceptable?   Dunno!  I think a hydraulic stretch test is the only way to find out what the breaking strain actually is, and as the risk can be avoided entirely by making male and female pitches identical, that’s probably what a pro would do.  Pitch error forbidden!
• On a precision tool, like a micrometer, pitch error would be unacceptable.   On a less precise thread like a lathe lead-screw, pitch error matters much. Ancient lathes with worn lead-screws and half-nuts still do good work!

My applications don’t stress the threads I make,  and on the rare occasions strength matters I buy the commercial product,  because these are much stronger than home-made.   (Rolled, not cut, and ground when accuracy matters.  And made from better steel.)

Thus far I haven’t worried about the pitch error inherent in my compromise Metric/Imperial Lathe, because strength isn’t critical.   I also make extensive use of Tubal Cain’s advice on reduced thread engagement as a way of reducing the strain on taps and dies.   Given doing that is OK for my purposes, I doubt any weakness due to my lathe’s pitch errors matter much.  Gut feel is that most amateur workshops are like mine in this respect, but could be wrong.

In other workshops, pitch error might well be critical, but having typed this post, I’ve veered to think being critical means the machinist has to do the sums, not trust a rule of thumb.

How much pitch error matters, if at all, depends on the size of the error and the circumstances.  As circumstances cover a huge range of requirements,  there may not be a simple rule of thumb.

For what it’s worth none of my books discuss allowable pitch error on threads.   They all assume male and female will be cut to the same pitch.   I don’t have a really good lathe book though, and one of them might do better.

Dave

[Author]

Posts