Metric threads on South Bend

[old mart]

It is the 127 gear that is essential, 127/100, or 127/120 or 127/125 can all be made to work. I happened upon a 127/125 which was fairly close on diameter to the standard 120 gear on the Smart & Brown model A. The new gear was metric Mod1, so all of the old imperial gears that we had, (only 3, the 120 and a couple of 50's) were mothballed and I bought metric ones and did the maths to have all the common pitches from 0.8 to 2.5mm. Using inch pitches is easy too, as the inputs only need to be matching sizes and any idler will do, it works equally with the 125 or the 127. The metric pitches are exact, or within 2% and this lathe uses a 6tpi leadscrew.

[J Tiers]

Thanks to everyone who shared their knowledge and experience on metric thread conversion. As I mentioned before my Grizzly lathe uses 91/86 compound gear for such conversion. Its accuracy is in between 47/37 and 80/63 combinations. ................................................

our lead screws may have bigger errors in themselves and trying to make the exact metric thread conversion is a loosing game.

When leadscrew errors are factored in, it may be a losing game even if you have a lathe made for metric (or imperial). The leadscrew errors are still there. If they are not given in the specifications, then you haven't any idea where you are regardless of what sort of conversion you use, or if you thread the native thread types for your lathe.

I rarely see leadscrew accuracy quoted. Even Monarch does not give that spec for the 10EE in the 1961 brochure I found, although it's a fair bet that the accuracy is very good indeed..

Another brochure quotes the 10EE specs for 1 foot and 4", which are 0.001" per foot, and 0.0004" per any 4" section. The "per foot" is only slightly better than the conversion using the 80/63 combination, although the 4" spec is much better than either "approximation". They say you probably will get better than that, due to their internal QC limits, but the published spec is all you can count on.

How much larger is the error in your lathe, or mine? We have no idea, but given the specs from Nook, I would bet on a 4 to 7 thou per foot error, which would be about 0.0003 to 0.0006" error per inch, as a minimum. The approximate conversions are both better than that.

Those 127 tooth gears are both a size nuisance, and expensive to buy. What price will you pay to get that part of the threading process to be exact? What will you do about the other errors?

It is the 127 gear that is essential, 127/100, or 127/120 or 127/125 can all be made to work. ....................... The metric pitches are exact, or within 2% and this lathe uses a 6tpi leadscrew.

If pitch within 2% is acceptable, why bother with the 127 tooth gear setup? Either of the other "approximations" does about 100x better.

[old mart]

The 2% was just a guess, the calculations are at the museum, probably less. But even with the 127 teeth, there was an error in half the results.

[Paul Alciatore]

Having produced and done a lot of research on thread conversion gearing, I can tell you there are two big reasons for using the approximate gear combinations. First, the 127 tooth gear is going to be big. For most lathes, it will be the biggest change gear in the entire collection: it certainly is in mine and I have more change gears than most. And big gears need more space on the banjo (bracket). Besides just needing longer arms on the banjo, some desired combinations of gears just will not fit. So smaller gears do have an advantage when building these gear trains on your lathe.

And, from the manufacturers' perspective and perhaps yours as well, making smaller gears costs less. Less in materials and less in time. So they are less expensive.

For the absolute purists, here's perhaps an even better solution:

An interesting thought hit me while reading all the comments about the accuracy of lathes' lead screws. First, I have never seen any accuracy specification on a lathe's lead screw. I am sure the OEMs do have them, but I have not seen them. And I have no idea as to how accurate or inaccurate the lead screw on my SB lathe is. But here is a thought on how you might improve on the accuracy in your particular case.

First measure the error in your lead screw. Chances are it will be better than the absolute spec for your lathe which may be in the vicinity 0.01%. But, for this, you mainly want to know which direction it is in, + or -.

Now, the 47/37 pair of gears gives a conversion ratio of 1.270270... which is about 0.00027 (~0.021%) on the high side while the 80/63 pair gives 1.269841... which is about 0.000159 (~0.012%) on the low side. So you have a choice of a positive or a negative error in the conversion ratio. And the errors in the lead screw and conversion ratios are similar in magnitude.

The 47/37 pair with a positive error will make the lead screw turn faster and the metric pitch will then be larger. So, if it is combined with a lead screw with a negative error, the two will partially cancel out and you may/probably will wind up with better accuracy than if you had used the 127/100 pair with that lead screw.

Likewise, the opposite is also true if your lead screw has a positive error and you use the 80/63 pair with it's negative error.

I did not explore it, but there are other pairs of gears that give even larger errors. They may be helpful if your lathe has a lead screw with a larger error. For instance, a lathe lead screw with a 0.5%, positive error (0.006" per foot) used in combination with a 48/38 pair of gears (1.26316... ratio or -0.5%) would be brought down to a 0.1% error. That would be a great improvement in it's accuracy.

I found that 48/38 pair by just adding one to both numerator and denominator of the 47/37 pair. And those gears may be a lot easier to either locate/buy or to make. In fact, I have a 48 tooth change gear in the standard set sold by SB.

Therefore, with a little trouble in actually measuring the error in your lead screw (gauge blocks or a good digital caliper) and with a little math, you can actually improve the accuracy of the metric threads that you cut.

[BCRider]

I noticed that a lot of the posts here seem to be struggling to compare the imperial standard of TPI to the lead per thread as used in metric.

I wonder if it would help if the thinking were to change to transferring the TPI into span per thread. So 20 TPI becomes .050" and 32TPI becomes .032". With this idea of comparing apples to apples I think the gearing changes will suddenly make more sense. Of course we also need to stop and thing about how many thousandths each metric pitch lead converts to so the ratio makes more sense.

[Hopefuldave]

Thats the real difference - Imperial is threads per length, Metric length per thread.

Holbrook (on my Model C, at least) got around this very elegantly in the QCGB: the leadscrew is driven by a sliding gear that can engage to gears on either of the shafts in the 'box, and the idler on the quadrant engages a 127 tooth for metric on the "cone" of gears, or a 45 tooth to the long gear the QCGB idler runs along, Simple.

The translations in the QCGB then give threads per length (imperial) one way, length (in mm) per thread the other, the numbers following the same progressions: 16,18, 19, 20, 22, 23, 24, 26, 28, 30 tpi, e.g., metric with a swap of the stud gear 1.6, 1.8, 1.9, 2.0, .22, etc, mm pitches, without the swap (and swinging a lever on the gear train drivng the stud gear, doubling or halving) 1.0, 1.125, 1.375, 1.4375, 1.5, 1.625, 1.75, 1.875 mm - some very non-standard threads there, but all accurate! Juggling levers it'll do anything you'd want, 60 tpi's from 2 to 120, 90 mm-pitches 0.2 to 15 , all accurately.

I think I've probably used about a dozen out of the lot... feeds are just as plentiful, I DO use many of those!

Dave H. (the other one)

[mikey553]

Therefore, with a little trouble in actually measuring the error in your lead screw (gauge blocks or a good digital caliper) and with a little math, you can actually improve the accuracy of the metric threads that you cut.

Measuring your composite pitch error (leadscrew plus gear ratio error) may not be that easy. Speaking of the thread I cut yesterday, it has a pitch of 1 mm. If we count exactly 100 revolutions of the spindle, the carriage should ideally move by 100 mm. The expected gear ratio error in my case is .0002 mm per revolution. So in 100 revs we expect to see .02 mm error. Not exactly very easy to measure.

If you have a good DRO on your lathe (I don't have one), you will be able to see something. Otherwise - forget it. Even counting the revolutions exactly is not an easy task. In addition every lead screw may have local and cumulative errors. And they are not distributed evenly along the screw length. Temperature deviations will play a big role in these measurements as well.

So I would suggest to not get too excited about exact metric thread conversion. In most cases we can do just fine without it.

[Paul Alciatore]

I did not say checking your lead screw was easy. I just tossed out a thought I had about obtaining greater accuracy. Frankly I doubt that it would be necessary or even noticeable in 99.99% of the threads any of us has or will ever cut.

[Baz]

Someone mentioned grizzly using 91/86. Are you sure as it isn't anywhere near 1.27? perhaps 91/76?
Also for information the early Myfords used 2x46/73
Following up this idea of pairs near to 1.27 maybe 70/50 (aka 35/25) is a crude option as they are already in a typical set.

[BCRider]

Someone mentioned grizzly using 91/86. Are you sure as it isn't anywhere near 1.27? perhaps 91/76?
Also for information the early Myfords used 2x46/73
Following up this idea of pairs near to 1.27 maybe 70/50 (aka 35/25) is a crude option as they are already in a typical set.

Mine isn't a Grizzly but it came from the same factory under a different distributor name. It has a 91/86 combo gear.

I see a lot of focus on trying to match the 2.54cm/inch or the 1.27:1 ratio idea. But that does not have to be the case.

So my 91/86 (checked and it is in fact 91/86) is a 1.058 ratio.

When using that ratio and the levers are set to C and 6 it is supposed to provide me a 1.0mm pitch. C6 in imperial with is 24 TPI. 24TPI gives us 1/24=0.0467" lead per thread. The 91/86 gear is inserted to provide an 86/91 reduction. Multiplying that by the 24TPI lead per thread of 0.0467" equals 0.0394" of lead per thread. And that works out to 1.00076mm according to my Windows conversion calculator.

But to add to this I've also got a 36T gear which is supposed to replace one of my 45T gears for some of the other lead values. So clearly the 86/91 setup by itself is not the be all and end all.

So there's more than one road to Rome on this quest for metric threading pitches from an imperial gear box. And it explains why there's a few different options.

[DennisCA]

This is what the 127/100 gear looks like on my SB9 clone

[J Tiers]

Looks like you can just barely close the cover. In other cases the 127 tooth gear prevents that.

[mikey553]

I did not say checking your lead screw was easy. I just tossed out a thought I had about obtaining greater accuracy. Frankly I doubt that it would be necessary or even noticeable in 99.99% of the threads any of us has or will ever cut.

Paul, I am not picking up on you, quite the opposite. I am very grateful for your extensive writeup with a theory behind the inch/metric conversion. Just wanted to clarify that accurate checking the lead screw for precision is probably beyond the average HSM capability.

[mikey553]

Someone mentioned grizzly using 91/86. Are you sure as it isn't anywhere near 1.27? perhaps 91/76?
Also for information the early Myfords used 2x46/73
Following up this idea of pairs near to 1.27 maybe 70/50 (aka 35/25) is a crude option as they are already in a typical set.

That was me who mentioned 91/86 gear on Grizzly lathe. It is not a mistake. This gear has a 91:86=1.0581395 ratio. Compare it to 127:120=1.058333 ratio. My gear has a theoretical error of .01828%, which is better than 47/37 gear (.021281% error), but worse than 80/63 gear (.012498% error). Look at Paul A. writeup above in post 25, the second gear in 127/x pair can have other number of teeth than 100.

[J Tiers]

.................................. Just wanted to clarify that accurate checking the lead screw for precision is probably beyond the average HSM capability.

That is undoubtedly true.

Accurately measuring to a few tenths over a longer distance, such as a foot, is not trivial. Give a bunch of actual paid machinists a 0-1" mic (or a 25mm mic) and a piece to measure to tenths (0.0025mm), and there will be a significant "scatter" in their results.

With a leadscrew, just a slight error in the position of the measuring tip relative to the centerline of the screw will throw the results off by a significant amount. The measurement is very sensitive to variables other than the actual distance being measured, due to many sources of error just in the setup used.

One might suggest that "well, the whole idea is to move the carriage, so we will use the leadscrew to move the carriage; problem solved!". That seems reasonable,and there are several good ways to do it. Using the DRO is one way to do it, if there is a suitably accurate DRO in place (not a given by any means). That avoids several errors that other methods might introduce.

But even that suffers from inaccuracy due to how the halfnuts close, whether they shift during the test, shifts in other parts, etc. It cannot separate out issues of slop in the carriage, lengthwise movement of the leadscrew in its bearings, etc. It's a good method of determining the general accuracy of effective carriage movement, but does not directly measure the leadscrew.

One can argue that it is the carriage movement that matters, and that is a good point. The extraneous errors would affect seem to affect threading also. But do they really? Threading is done with a back pressure on the leadscrew. Without a similar back pressure, some errors are not measured, and others are allowed that might be avoided.

It's an interesting problem. And probably irrelevant for most of us.

Our leadscrews are not up to the standards of the 10EE, even if we own a 10EE. If we do not, then the leadscrew is likely not ground to the precision of Monarch's production, and will probably be in error by several thou per foot. If we DO own a 10EE, then the leadscrew WAS accurate when made, but has had decades of wear since.

That "several thou per foot" may not be so bad. An error of 5 thou per foot is 5/12,000, or 0.04%. Not so bad, actually.