Making linear actuators? Possibly using brushless motors?

[Peter "Sci" Turpin]

Anyone got any ideas on this?

I'm aware of the servo mod version that's essentially putting some
threaded rod on a full-rotation servo. It's slow and has no position
feedback.

I'm aware there are commercial ones of similar scale to regular servos,
but they're rather expensive (seem to start at �70 per servo).

Wondering about direct-driving a threaded rod with a low-friction nut
(delrin, phospor bronze) with a small cheap outrunner brushless motor.
They're faster and with better torque than brushed DC motors. If the
thread pitch isn't too course, could likely move something with
reasonable force at very good speed, and probably a lot quieter than a
regular servo. Or at least it seems to me it would.

Anyone with more experience of these things want to chime in?

Sci

[benjamin winston]

Get a sensor-based brushless design, since they have the torque right
from the get-go. Sensorless ones first have to do phase detection
meaning they are 'jerky' off the line while they figure out the
position of the spindle.

Also, as side considerations, have you considered a) making one just
using a solenoid and b) making a pneumatic one.

b

[Glen]

I do this all the time, you don't even need a low-friction nut, BZP steel nuts on BZP studding works just fine.
There's two I made above the Hackspace's beer fridge, in the Reprap.

On Sunday, 3 June 2012 00:59:16 UTC+1, Sci wrote:

Anyone got any ideas on this?

I'm aware of the servo mod version that's essentially putting some
threaded rod on a full-rotation servo. It's slow and has no position
feedback.

I'm aware there are commercial ones of similar scale to regular servos,

but they're rather expensive (seem to start at �70 per servo).

[Simon Howes]

Whats your application?

I'd say you're best of with a stepper direct coupled to a leadscrew as its inexpensive and existing (multi axis) stepper driver boards can be had for peanuts nowadays on ebay. It requires a little faith as you command it to move x steps and hope it gets there. If you need some sort of feedback add an optical or mechanical encoder to the other side of the stepper. Use this just for stall and reset to position orientation, dont try to drive using this in your pulse train code.

Again, your application will specify your needs. Inexpensive RC servos can be good for all sorts of things but the cheap ones only seem to implement the P of the PID. This means they'll happily put out a small proportional amount of power and sit just shy of your commanded position getting warm, but without the I they'll never get there. This isnt a problem in some animatronic rigs.

DC motors can be good - I have some on my CNC Mill, but they dont have high low speed torque so you need to gear them down. And the drivers/optical encoders and even the motors are damned expensive.  Steppers dont suffer this and are very cheap nowadays, but dont come with integral position feedback (though as mentioned you can readily add a lower resolution approximation/stall sensor. Some cnc packages like mach3 even support this)

Which again all depends on what you're making. :)

[Peter "Sci" Turpin]

Thanks for that, will remember the bit about phase detection.

Solenoids only really have two position states (and I think are
comparatively weak without a fair bit of current), and pneumatics need
compressors and storage tanks, and are fairly noisey.

[Peter "Sci" Turpin]

On 03/06/2012 12:38, Simon Howes wrote:
> Whats your application?

Looking into options for final version of a Stewart platform for a
robotics project. Will likely use servos for prototyping, but I think I
really want something with better range and less noise for the final
version, hence looking into making suitable linear actuators.

> I'd say you're best of with a stepper direct coupled to a leadscrew as
> its inexpensive and existing (multi axis) stepper driver boards can be
> had for peanuts nowadays on ebay. It requires a little faith as you
> command it to move x steps and hope it gets there. If you need some sort
> of feedback add an optical or mechanical encoder to the other side of
> the stepper. Use this just for stall and reset to position orientation,
> dont try to drive using this in your pulse train code.

Steppers are indeed another option, and I do have a lot to hand, but
it's hard to find size/power comparisons for them vs DC outrunners
(outrunners noted for having much better torque than inrunners and
regular DC).
So for a motor of the same approximate size/volume, how would outrunners
compare with steppers? I mean since an inrunner DC is basically a
3-phase DC motor, and a stepper is a 4-phase DC motor. Stepping would
obviously be a lot more precise, but DC has the speed which would be
favourable for a fine-pitch leadscrew drive.

> Again, your application will specify your needs. Inexpensive RC servos
> can be good for all sorts of things but the cheap ones only seem to
> implement the P of the PID. This means they'll happily put out a small
> proportional amount of power and sit just shy of your commanded position
> getting warm, but without the I they'll never get there. This isnt a
> problem in some animatronic rigs.
>
> DC motors can be good - I have some on my CNC Mill, but they dont have
> high low speed torque so you need to gear them down. And the
> drivers/optical encoders and even the motors are damned expensive.
> Steppers dont suffer this and are very cheap nowadays, but dont come
> with integral position feedback (though as mentioned you can readily add
> a lower resolution approximation/stall sensor. Some cnc packages like
> mach3 even support this)

Depends on the sort of DC motors though. From what I gather they go in
order of torque; Brushed DC < Brushless inrunner < Brushless outrunner

But where do steppers go in this order?

Another option is cost. I have a lot of steppers on hand (about 2000
actually, but fairly weak ones), but if I didn't how would a motor of
similar power compare in price with the �5 for a 160Watt brushless
outrunner. That would possibly need gearing down tho, so increases
mechanical complexity. Depends on leadscrew pitch.

Also position on a linear drive might well be easier. Depending on
length, a linear pot or a linear optical encoder. The latter should just
need a laser printer and some clear acetate to print on.

> Which again all depends on what you're making. :)
>
> On Jun 3, 2012 12:59 AM, "Peter "Sci" Turpin" <s...@sci-fi-fox.com

[Peter "Sci" Turpin]

I'd like to consider how the best possible way of doing things might be
before stepping back to likely cheaper & simpler options.
Standard M8 thread will work, but it'll have backlash and bursting force
on the nut, as well as being less mechanically efficient than using a
square or acme thread. More efficiency means more performance from the
motors used and less noise.

Since square thread is very pricey, thinking 1/4" acme thread may be a
good one. Getting regular M thread in sub-1mm pitches gets to a similar
price point.

[Adrian Godwin]

I was also tempted by brushless for a linear actuator beacuse they're
so cheap and mechanically simple, but as ben says, the sensorless ones
are a problem at low speed (and sensor ones are expensive). You
probably need low speed for precise positioning, so that's tricky. To
get the sensorless ones working, you need enough speed to generate a
measurable back-emf from the unpowered phase (and you also need
back-emf on the powered phases to keep the current down).

It's possible that if you can measure the linear position accurately
enough, and have low backlash, that you could infer the rotational
position from the linear position. But I think you'd need a hefty
current to get decent torque, and you need linear control of the
currents in each branch unless your gearing is low enough that 1/3 of
a turn corresponds to your minimum step size.

Ideally, you want ballscrews for low backlash and friction. They're
expensive - but it's actually the nuts that are expensive, not the
rolled screws. So they might be a good alternative to acme if we could
devise a cheap, low friction nut. Cutting one from ptfe might be good
enough.

-adrian

[Adrian Godwin]

On Sun, Jun 3, 2012 at 9:30 PM, Adrian Godwin <artg...@gmail.com> wrote:

> Ideally, you want ballscrews for low backlash and friction. They're
> expensive - but it's actually the nuts that are expensive, not the
> rolled screws. So they might be a good alternative to acme if we could
> devise a cheap, low friction nut. Cutting one from ptfe might be good
> enough.
>

Or casting one from molybdenum-loaded epoxy (moglice), come to think
of it. cast in place over the ballscrew itself.

-adrian

[Adrian Godwin]

Here's an extreme approach of this method, essentially a brushless
motor integrated with a ballscrew to make a drop-in replacement for a
pneumatic actuator.

http://www.eurekamagazine.co.uk/article/3006/Novel-linear-motor-offers-alternative-to-pneumatics.aspx

[Peter "Sci" Turpin]

Reading the article it sounds more like a digital solenoid than
ballscrew, but it's an interesting method.

[Peter "Sci" Turpin]

Would have thought just adding your own hall-sensor to the motor would
be easier and more accurate than trying to infer the motor position by
the linear motion.

With cheap motors not having much low-end torque, that either leaves
gearing them (with noise & efficiency losses) or switching to steppers
which have a much straighter torque-curve. Sounding increasingly like
it'd be the best option.

Just have to see how fast I can drive small steppers.

On 03/06/2012 21:30, Adrian Godwin wrote:
> I was also tempted by brushless for a linear actuator beacuse they're
> so cheap and mechanically simple, but as ben says, the sensorless ones
> are a problem at low speed (and sensor ones are expensive). You
> probably need low speed for precise positioning, so that's tricky. To
> get the sensorless ones working, you need enough speed to generate a
> measurable back-emf from the unpowered phase (and you also need
> back-emf on the powered phases to keep the current down).
>
> It's possible that if you can measure the linear position accurately
> enough, and have low backlash, that you could infer the rotational
> position from the linear position. But I think you'd need a hefty
> current to get decent torque, and you need linear control of the
> currents in each branch unless your gearing is low enough that 1/3 of
> a turn corresponds to your minimum step size.
>
> Ideally, you want ballscrews for low backlash and friction. They're
> expensive - but it's actually the nuts that are expensive, not the
> rolled screws. So they might be a good alternative to acme if we could
> devise a cheap, low friction nut. Cutting one from ptfe might be good
> enough.
>
> -adrian
>
>
> On Sun, Jun 3, 2012 at 6:11 PM, Peter "Sci" Turpin<s...@sci-fi-fox.com> wrote:
>> I'd like to consider how the best possible way of doing things might be

>> before stepping back to likely cheaper& simpler options.

>> Standard M8 thread will work, but it'll have backlash and bursting force on
>> the nut, as well as being less mechanically efficient than using a square or
>> acme thread. More efficiency means more performance from the motors used and
>> less noise.
>>
>> Since square thread is very pricey, thinking 1/4" acme thread may be a good
>> one. Getting regular M thread in sub-1mm pitches gets to a similar price
>> point.
>>
>>
>> On 03/06/2012 11:36, Glen wrote:
>>>
>>> I do this all the time, you don't even need a low-friction nut, BZP
>>> steel nuts on BZP studding works just fine.
>>> There's two I made above the Hackspace's beer fridge, in the Reprap.
>>>
>>> On Sunday, 3 June 2012 00:59:16 UTC+1, Sci wrote:
>>>
>>> Anyone got any ideas on this?
>>>
>>> I'm aware of the servo mod version that's essentially putting some
>>> threaded rod on a full-rotation servo. It's slow and has no position
>>> feedback.
>>>
>>> I'm aware there are commercial ones of similar scale to regular servos,

>>> but they're rather expensive (seem to start at �70 per servo).