Re: [CAD_CAM_EDM_DRO] cleaning, lovejoy, black box tach, linear scales, making pulleys

jmw@... wrote:

> 2) I just picked up some new / surplus helical connectors. These
> things do wind-up; hold one end in one hand, the other end in the
> other and tweak. I din't put the thing in a fixture, use an oz-in
> torque wrench, etc--but I could feel the wind-up without twisting all
> that hard. I bet that with the right choice of (urethane) elastomer
> for the spiders, Lovejoys will work just as well for low inertia
> loads.

That's why you use helical slit couplings for shaft encoders, but not,
generally, for servo motors. The low drag and inertia of the couplers
is no problem there.

> 3) In answers to a previous post I learned that both tachs and
> encoders are used in servo systems because at one time processor
> speed was too slow to calculate motor speed from the encoder output.

No, the real reason is there is no feedback between pulses from the
encoder. When a pulse comes in from the encoder, you don't know
whether this is due to a very slow creeping of the servo, or a sudden
impulse from the cutting tool biting into the workpiece. You would need
several counts from the encoder to distinguish the two. The DC tach,
however gives continuous feedback of velocity info, you don't have to
wait for encoder counts at all.

> But why do tachs persist? I understand that certain amplifiers and
> control boards want a separate tach input--but wouldn't it be easier
> to make make a black box which takes encoder input, calculates RPM
> and outputs x volts per 1000 RPM as wanted by other components of the
> system. This sounds easier, more reliable, less fussy and maybe even
> cheaper than mounting 2, 3 or 4 mechanical tachs. But this doesn't
> seem to be current practice ...

The problem is that any information derived from an encoder is
time-discontinuous, whereas a DC tach gives a time-continuous
response. In other words, the time lag for the DC tach is constant,
but the time lag for information from the encoder varies with
velocity, making filtering and servo loop compensation more difficult.

> 4) If fast responding, high resolution linear scales existed, isn't
> it the case that (at least for closed loop systems) backlash, ball
> screws, helical connectors, etc would all be irrelevant?

Yes, and they do exist. But, most home shop types and CNC
self-builders can't afford them. You could spend $100K for a
truly fast responding, high res linear encoder of machine tool size.
As soon as you get past .0002" / count, the prices get astronomical!

> Trying to
> measure position indirectly from a rotary encoder seems like the root
> of all evil. Maybe really high-grade scales wouldn't seem so pricey
> if you didn't have to retrofit ball screws. I understand there's a
> problem with table movement within the backlash zone, but why can't
> the motion controller mimic a human machinist's habits of taking up
> backlash when reversing direction?

Because a human machinist, or a servo system, can't get from one side
of the backlash to the other side instantly. So, you can't have a smooth
reversal, like when interpolating a circle.

> 5) Somewhere (maybe in this group) I read a suggestion about making
> HTD pulleys by drilling holes around the desired OD and then taking a
> clean-up cut on the lathe. Flanges could be screwed on ex post? Has
> anyone done this? Unless I'm reading the price lists altogether
> wrong, an afternoon spent this way could be quite rewarding. Of
> course you'd go mad without a rotary table or index setup.

Yes, but the circular arc section for the tooth face is not the right
shape for most toothed belts. I guess the HTD belts have rounded
'teeth', but they probably have a greater tendency to climb the cogs
under high load.

Jon