Hobbing a rack?

You will have to set the hob over at its lead angle and feed at the rate of 1 pitch of the hob/rack per turn of the hob, however the resulting rack will have a scalloped profile when viewed from the end as you have no feed across the rack

Peter>

You certainly have to do that. However, to get a full cut across the rack you would need to do multiple passes, stepping a few thou across the width of the rack with each pass.

An alternative method (although not using a hob) is to strap the rack blank to the outside of a (large) cylinder & screw cut it...if the cylinder diameter is large relative to the tooth pitch the helix angle is so small that it doesn't matter. This is how the racks for the Taig lathe are made...they have a cylindrical fixture that can hold several racks (8? 12?) & they are all screw cit at the same time.

Regards, Tony

Nice to finally shake your hand at the Midlands recently.

Yes, I'd considered that and one would need to have an accurate backwards movement if hobbing electronically.

Talking of which, does your Division Master work with an incoming pulse train for gear hobbing synchronisation?

(Not for me, I'll generate my own from the SSI TTL left over from my computer project (own instruction set) from 40 years ago)

(Rack is needed for a Ham Radio RX, for a "Catacomb" style bandswitching)

Indeed - good to meet you after all this time.

You can drive the DivisionMaster in "slave mode" from an external step signal - but of course, at that point, there is no advantage in using the DM vs using a separate stepper drive, unless the DM is the only spare stepper driver that hou have to hand. You can also signal to the DM to tell it to step to its next division, but I'm not sure that helps much in this case.

Regards, Tony

Had you ever considered driving the hob from a stepper as well (Possibly with fly wheel attached to reduce the jitter); in which case, sybchronising the two steppers (hob, workpiece) would be very easy by dividing down appropriately from a common master clock?

Certainly considered it - and yes, you may well need to use a flywheel of some kind, but more importantly, you would probably need to gear the motor down a fair bit to get adequate torque to drive the hob. (Or an erfing great stepper motor...). I bought a couple of toothed belt pulleys a while back with the intent of experimenting along those lines, but nothing concrete so far.

Dividing down from a common clock to 2 stepper drives is potentially an interesting solution - there is a bit of mathematical juggling to do to ensure that you handle the rounding errors properly, but that isn't fundamentally any more difficult than the maths that I already do in the DM to handle rounding errors. The alternative approach is to attach an encoder with a suitable fine resolution to the hob spindle & divide down from that - which is essentially what Brian Thompson did for his hardware-only hobbing controller (published in MEW a few years back) - make the pulse-per-rev count on the hobber high enough and you just divide down to get the pulse traing for rotating the blank. Problem with that route is that you need a high count encoder, or a low count encoder plus a suitable gearing up to drive it faster than the hob spindle.

Regards, Tony

A few thought experiments ...

But if you start with a high enough master clock, then there is no essential difference between the two approaches (that and a fine resolution optical disk)

... or a low-count encoder acting as the frequency reference for a Phase Locked Loop master oscillator.

But even so, in the case of rounding errors, surely they'd even out (or oscillate slightly about a mean) in the case of simple |(ie, non computed) TTL division chains, because unlike the assumed calculation in your DM, where you'd be working out the count to the next division, in the case of hobbing, all oscillators / optical disks / division are running continuously?

In respect of driving the hob, Giles Parkes has suggested (since I first posted here) that the hob should be rotated at about 180 RPM, 3R Per Sec, which would be manageable with a direct drive stepper, but geared down might result in a risk of the stepper stalling?

Other thought - the stepper is AIUI a synchronous motor that has been adapted for stepping, so driving with high frequency quadrature sine waves would remove any jitter.

... and don't remember if I posted this here; I've no connection with these people; and download at your peril; I downloaded the Beta 2.0 version for Windows ...

PS. As well as Brian Thompson, there was a similar article in last Sept's MEW.

... forgot to add; if you have, say, a 40:1 reduction on the workpiece spindle (which is a mechnical divide by 40), then in the drive to the hob, you include an electronic divide by 40, then, in the first instance, the hob and workpiece will rotate at the same speed.

All that is then needed is a simple divide chain in the workpiece control to set the number of teeth (which, ridiculously and mysteriously, would also work for a two tooth gear :-) )

Thinking on my feet!!!! The way to go is now clear; a conventional motor driving the hob (probably the vertical head on the mill), with an optical disk having the same number of slots as the stepper motor has steps per revolution; a phase locked loop synchronised to the optical disk, but at a rate 40 times higher in frequency (assuming a 40:1 worm in the dividing head), then divided down by a simple TTL divider chain to give the number of teeth.

I've been throwing around some of these ideas recently, but how often the solution presents itself, when you explain your nascent ideas to another!

(A problem shared is a problem halved)

I am currently in Spain for two months but wish I was at home with the lathe and mill !!! If I was I would be trying the above problem on my lathe with ELS. The hob in the chuck and a blank (to end up as a worm wheel) driven by a stepper replacing the leadscrew stepper. I need to make a worm and wheel for a dividing head and I think this may be an option. I was considering using a 5/8th UNC tap as a hob. I know that Tony has an ELS, do you consider this an option?

Richard

It is certainly an option that would be worth trying. What isn't clear to me is how well it would work in practice, given that ELS only allows use of a 1 pulse per rev encoder - i.e., it is basically down to how well ELS will track the spindle speed & adjust the workpiece rotation accordingly. Which reminds me, I have an updated ELS ROM that improves the speed tracking algorithm - I haven't yet checked out whether it works...so many toys, so little time..

Regards, Tony

That is true.

Depends how the "running continuously" is achieved. If you are working from a master oscillator that conveniently runs at the (micro)step frequency needed to drive the stepper for the hob, then it is a simple integer divide, with no rounding errors, to get to the step rate for the motor driving the workpiece. If you have direct drive to both axes, it is all very simple; divide the hob pulse rate by 2 and you get 2 teeth, 3 gets 3 teeth...etc.

If you decide that you need a reduction drive to either axis, as long as you choose the same ratio for each, it is still nice and simple.

The problems come when you have different drive ratios on each axis - say, for example, you decide that you need a 7:1 reduction to get enough "grunt" out of the hob stepper to stop it stalling, and you strap your gear blank onto your 40:1 dividing head. Now, for each pulse on the hobber motor, you need 40/(7*T) pulses on the dividing head, which is not a whole number of pulses. So you have to make sure that the rounding errors don't accumulate over time, otherwise all you achieve is a complicated method for reducing the diameter of a gear blank.

What I had in mind was using gearing to increase the torque at the hob

- i.e., small gear on stepper, large gear on hob spindle, which would reduce the risk of stalling (but would also need the stepper to run faster).

That is essentially what microstepping aims to achieve - the sequence of phase current settings for each microstep approximates to a sine wave. (Not being of the analogue circuitry persuasion, I tend to think in terms of pulse trains rather than sine waves ;-) )

Regards, Tony

Yes.

Exactly so (see my answers to your previous post). And it could just as easily do a 1-tooth wheel ;-)

That is basically the Brian Thompson solution, except that he ups the encoder count by the factor of 40 so that you can use the pulse train out of the encoder direct rather than having to set up a separate PLL. John Stevenson has that setup - in order to use a (commercial) optical encoder that he had to hand, he introduced a gear train between the hob spindle and the encoder to increase the number of encoder puulses per spindle rev by the appropriate ratio. I guess that is just a mechanical PLL though ;-)

Regards, Tony

Not quite sure I follow this, but..

..if the problem is that the encoder only gives one pulse per revolution, why not use one which gives more? They are very easily and cheaply available in mice - the one I just took apart has 40 pulses per rev, but ymmv.

-- Peter Fairbrother.

Yes, that is indeed one way to skin this cat.

Regards, Tony

Another thing I have done is to use the led/transistor pair unit from a mouse and then make my own wheels - useful if eg you need a wheel with a

40mm hole in the center to go on a lathe spindle, or whatever. Make the lines and spaces about the same size as the lines in the mouse wheel - this will mean a few hundred lines on a large wheel.

To make the wheels I use a vector drawing program (inkscape) and then laser print them onto transparent film. Make VERY sure you use laser-compatible film, damhikt. I have then had the film laminated for extra strength, but that is not always necessary.

Mice can be very useful - I buy 3 at a time for 99p each delivered from China (ebay), one will be put to work and two get used for parts. Three small microswitches, three optical wheels/sensors, plus a USB cord and some other bits per mouse - well worth it.

-- Peter F

No, much simpler, assuming that both steppers are of the same resolution, be it 200 steps or 48 steps or whatever, if one side has a mechanical division by gears or by worm and wheel, then in the other side insert a fixed electronic division of the same ratio.

What may make my whole proposal fall flat on its face is the problem of steppers stalling at high pulse rates.

Yes I know that the software was changed due to the "long thread problem". I have yet to update my pic. I will certainly give hobbing a try when I get home. My gut feeling is that a 3Nm stepper with a aluminium blank just bored and grub screwed to the shaft will probably work for a first try. When screw cutting I have never seen any speed change whilst cutting. As long as the hobbing cuts are light I feel things should work ok.

Obviously born in Yorkshire or Scotland! Makes a lot of sense to me even when born in Gloucestershire!

I certainly noticed some speed fluctuation when screwcutting on my Myford - shoulda gone for the bigger motor when I fitted the VFD - but nothing that couldn't be fixed by sticking it in back-gear.

Regards, Tony