Gear hobber (mechanical)

I'm looking for information about the gear hobber of college engineering.My special interest is pointed to the production of prime number gears, for example the 127 teeth gear. Now i want to know, whether the kit or the plans form college engineering will give enough information about the construction of the differential gear addition to the jacobs gear hobber.

My email request to college engineering to answer this question was not responded until know.

Is there any member of the forum to give me more information?

I assume you are aware of the modified version in the USA by Gary Martin Models. maybe they can supply information.

I'm not aware of the use of a differential in these machines so interested to hear what you currently know about it.

Do you particularly want to hob gears, of just make prime number gears? These days, I would have thought that a stepper driven dividing head, using for example Steve Ward's controller, would be much easier.

http://www.worldofward.com/rotarycontroller/overview/

Gustav,

You will find Mike Haughton (MEW Author) has a lot of experience on Jacobs Hobber, if you PM me with your email I will forward your queries.

FYI.

I've looked at the Martin Models castings (on line) for their gear hobber.

But - a Brit (Andy Pugh) has done interesting things gear hobbing with LinuxCNC (old name of software: EMC). Go to Youtube, and look for "Andy Pugh Gear Hobbing" and see what comes up.

(This might be much easier/quicker to build than a mechanical hobber, but I *do* like the Martin Models version...)

John.

Many thanks for the posts. I hope that Mike Haughton will help with information. I'm more interested in the old- fashioned way with a gear train, i think that I'm to old for the electronical way of construction.

I would have thought the simplest and easiest way to make a 127 and others would be to use a dividing head with yet another worm and wheel driving the worm on it as per the G Thomas design of old.

In some ways it's better than differential dividing as the flank errors can be worked out and then averaged rather than getting as close as possible with summed fractions.

Must add though that some people desire things like 127T gears for conversion sets to allow metric to be cut on imperial machines. There is no need for them as approximations can be derived which exceed the accuracy that's likely to be achieved anyway even with it.

Stepper generally have a step accuracy of 5% by the way. I suppose that the usual argument can be made that this is reduced by the ratio of the worm and wheel they drive but I reckon dividing plates are better.

John

-

Edited By Ajohnw on 21/02/2016 23:10:04

"In some ways it's better than differential dividing as the flank errors can be worked out and then averaged rather than getting as close as possible with summed fractions."

I think differential indexing is exact but compound indexing uses the sum of fractions which cannot therefore create prime divisions.

You are correct Bazyle. No matter people probably understood anyway. I should go to bed rather than looking to see what's going on here.

In terms of change wheels people have generated very precise approximation for metric on imperial but I can't recollect seeing any for imperial on metric.

John

-

.

I'm not arguing, John ... but, on a typical 200 step/rev motor, that's one part in 4,000 [which probably compares favourably with much of what we do].

MichaelG.

Gustav

Just be aware that if you do decide to build a gear hobbing machine, the small hobs are not that easy to find these days.

I have a good selection for my own Mikron 112 hobber, but it took a good few years to obtain them and I still look every day on the off chance something I need turns up.

Phil.

I looked at in terms of angular error Michael. Initially because when they first appeared I thought what a simple dividing device used all on it's own with some form of zero shift clamp. I'd need to check again but think it's +/- 5%. There are better, a few cropped up on homeworkshop but went very quickly. Less steps and off a cnc machine if I remember correctly.

The Thomas design uses a 60T and 40T worm and wheel. That's 1 part in 2,400. Can't remember how many divisions on the dial but rather than make it on it, which is possible ones with 100 are about giving 1 part in 240,000. Good enough to make a sizeable gear but fiddly. It's a case of advancing so many divisions and periodically minimising the error by altering that slightly.

I did some sums before building one and then bought a Dore Westbury which came with the Westbury dividing head. I think the dial I had found had either 200 or 150 divisions but his design was ok. I do intend to fit a larger worm wheel to the Westbury one when and if I need to. Bigger plates too but for my limited use of it the items that came have been fine. Thomas describes how to make the plates in his book but a local supplier had a lot made and sold them at rather low cost. It wasn't worth even buying the blanks.

John

-

Edited By Ajohnw on 22/02/2016 15:18:29

.

Have a look at information from companies like Oriental Motor, & remember that the quoted error is [of course] non-cumulative.

Micro-Stepping does not help much, because it the intermediate micro-steps are not true detents ... but motors are available with 800 real steps/rev.

MichaelG.

127 tooth, 1 MOD.

The mechanical hobbing machine looks fun.  I agree, though---won't you then have to make the hobs?  So that leads you into making the Eureka relieving tool.  How far do you want to take this?

Edited By MyrtleLake on 22/02/2016 17:37:03

If I have done my sums correctly a 200 step driving a 90T worm, 5% is equal to around +/- 0.001" error per inch of radius.

Cranky old calculator. VPAM drives me up the wall but I'm sure the calc acts up now and again, I even liked reverse polish, even less key strokes.

Loosely speaking 800 steps would 1/4 that if it's still 5%. That's more like it.

John

-

The Sherline rotary table...

400 steps per handle revolution. Worm and gear makes 72 turns for one full rotation of the table. So 28,800 steps per 360 degrees. That equates to 80 steps per degree, which in turn equates to 0.75 minutes per step. Final controller, electronic position is accurate to +/- 0.5 step.

"This turns out to be a very small amount because the sine of this angle is only 0.0001091. This will amount to 0.109 mm per 1 meter when compared to a flat plane. This is just 0.0055 mm at the edge of the table"

http://www.sherline.com/8700inst.pdf

Good enough for me!

So what happens to all the calculations if your plates are generated on a stepper driven CNC ?

Proper CNC machine I would hope John. The few I have bothered to look at in detail - older used one seem to use low step count motors. I assume like the far more accurate ones I mentioned. So far I have resisted buying one.

As to sin and pdf. I used tan but the answer is virtually the same anyway. Really a different type of trig should be uses as sin and tan don't equate to arcs.

360/ (90*200) = 0.02 degrees, 5% of that is 0.001 degrees. Sin of that 0.0009999999 tan 0.001000. So both come out at one thou per inch of rad. The assumption that 800 steps would 1/4 this is fairly valid if the error is the same. 400 would halve it. A 72T worm would make it worse.

That sort of error with 200 pulses is a bit debatable on larger gears such as 127T. As Michael pointed out micro stepping is more iffy and steps will be made anyway. Where high accuracy is needed on things like astro telescopes the answer is generally a lot of gearing and closed loop feedback - the star is tracked and the steps corrected as a result. The other option used is to fit an encoder and drive according to that - not step counts. Some do the obvious then and use an ordinary motor to drive instead.

One aspect of cnc is interesting. When I visited chester m/c tools the sales man pointed out that numbers come out of their simple digital scales and he wondered why some one hadn't done a machine that used these.

It could be done with +/- 1u scales too. The real ones of course if they exist.

John

-

Edited By Ajohnw on 22/02/2016 20:31:51

Bedtime but this site explains the problem rather well -

**LINK**

I often hear, exaggerated by me of course of some set up with a lead screw being micro stepped at 1/10,000 of a step so the system is accurate to blah blah when it wont be.

There are some rotary encoders about that offer a high power of 2 resolution as well but it's best to look at the claimed accuracy rather than the headline figure.

John

-