magnetic bearing

[colin]

Hi,

I think I might try making one of these,
basically I need more radial precision at each end,
to keep the opto encoders happy,
with the center ball race in the motor in the center of the shaft keeping
axial movement in check.
or maybe have one with radial controll there too.

if I use an old 40w transformer core
cut the E parts in half so I have U parts
I could make 4 or 3 solenoids and use a pic to generate PWM drive
and sense the resultant AC current via adc to determine inductance and hence
rotor position.

12mm hardened steel shaft might be able to hold this directly.
theres no load apart from the shaft.
but im not realy sure what size or power il need.
maybe ill need a bearing surface for when its powered down suddenly or
ensure there is a backup supply for the bearings.

and how quick would I need the control loop to be ?
1ms might be good to aim for.

would a PIC be upto doing it digitally like this ?
or do I need a faster linear system ?
I have a dspic33 already but its totally comitted.

I could suck it and see but im constantly doing this and although it can be
quite fun,
ive been spending too much time and energy on this so far and ive not got
much of either to spare atm.
shouldnt be too hard but would be nice to get it right first time.

there doesnt seem to be much in the way of off the shelf types,
although ive found quite a few products wich have them so far.

Colin =^.^=

 

[D from BC]

Hi,

I think I might try making one of these,
basically I need more radial precision at each end,
to keep the opto encoders happy,
with the center ball race in the motor in the center of the shaft keeping
axial movement in check.
or maybe have one with radial controll there too.

if I use an old 40w transformer core
cut the E parts in half so I have U parts
I could make 4 or 3 solenoids and use a pic to generate PWM drive
and sense the resultant AC current via adc to determine inductance and hence
rotor position.

12mm hardened steel shaft might be able to hold this directly.
theres no load apart from the shaft.
but im not realy sure what size or power il need.
maybe ill need a bearing surface for when its powered down suddenly or
ensure there is a backup supply for the bearings.

and how quick would I need the control loop to be ?
1ms might be good to aim for.

would a PIC be upto doing it digitally like this ?
or do I need a faster linear system ?
I have a dspic33 already but its totally comitted.

I could suck it and see but im constantly doing this and although it can be
quite fun,
ive been spending too much time and energy on this so far and ive not got
much of either to spare atm.
shouldnt be too hard but would be nice to get it right first time.

there doesnt seem to be much in the way of off the shelf types,
although ive found quite a few products wich have them so far.

Colin =^.^=

Is it me or is this wayyy over my head?
I think the coffee hasn't kicked in yet..
D from BC

 

[colin]

Is it me or is this wayyy over my head?
I think the coffee hasn't kicked in yet..
D from BC

Its just magnetic levitation basically.

Colin =^.^=

 

[Tim Wescott]

Hi,

I think I might try making one of these,
basically I need more radial precision at each end,
to keep the opto encoders happy,
with the center ball race in the motor in the center of the shaft keeping
axial movement in check.
or maybe have one with radial controll there too.

if I use an old 40w transformer core
cut the E parts in half so I have U parts
I could make 4 or 3 solenoids and use a pic to generate PWM drive
and sense the resultant AC current via adc to determine inductance and hence
rotor position.

12mm hardened steel shaft might be able to hold this directly.
theres no load apart from the shaft.
but im not realy sure what size or power il need.
maybe ill need a bearing surface for when its powered down suddenly or
ensure there is a backup supply for the bearings.

and how quick would I need the control loop to be ?
1ms might be good to aim for.

would a PIC be upto doing it digitally like this ?
or do I need a faster linear system ?
I have a dspic33 already but its totally comitted.

I could suck it and see but im constantly doing this and although it can be
quite fun,
ive been spending too much time and energy on this so far and ive not got
much of either to spare atm.
shouldnt be too hard but would be nice to get it right first time.

there doesnt seem to be much in the way of off the shelf types,
although ive found quite a few products wich have them so far.

If your goal is rigidity I don't think you're going to get there with
magnetic bearings. You may be able to get the average measured error
down to zero, but if your shaft experiences any vibration it'll wobble,
and that'll show up as a wobble or wow in your position measurement in
your encoders.

If your real problem is the sensitivity of your angular measurement to
your shaft axis position, why don't you address that? A resolver,
properly built, should be significantly less sensitive to shaft run-out
issues than an encoder.

Should you be determined to stick with an encoder, why not take the same
shaft position measurements that you would have made for your magnetic
bearings, and just use them directly to correct the encoder output?

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Do you need to implement control loops in software?
"Applied Control Theory for Embedded Systems" gives you just what it says.
See details at http://www.wescottdesign.com/actfes/actfes.html

 

[Chris Jones]

Hi,

I think I might try making one of these,
basically I need more radial precision at each end,
to keep the opto encoders happy,
with the center ball race in the motor in the center of the shaft keeping
axial movement in check.
or maybe have one with radial controll there too.

if I use an old 40w transformer core
cut the E parts in half so I have U parts
I could make 4 or 3 solenoids and use a pic to generate PWM drive
and sense the resultant AC current via adc to determine inductance and
hence rotor position.

12mm hardened steel shaft might be able to hold this directly.
theres no load apart from the shaft.
but im not realy sure what size or power il need.
maybe ill need a bearing surface for when its powered down suddenly or
ensure there is a backup supply for the bearings.

and how quick would I need the control loop to be ?
1ms might be good to aim for.

would a PIC be upto doing it digitally like this ?
or do I need a faster linear system ?
I have a dspic33 already but its totally comitted.

I could suck it and see but im constantly doing this and although it can
be quite fun,
ive been spending too much time and energy on this so far and ive not got
much of either to spare atm.
shouldnt be too hard but would be nice to get it right first time.

there doesnt seem to be much in the way of off the shelf types,
although ive found quite a few products wich have them so far.

Colin =^.^=

Here's a magnetic bearing web page that I found interesting:
http://www.amsat.org/amsat/sats/phase3d/wheels/index.html

Magnetic bearings would be great for long-lasting fans, I think. The
position of the rotor could be adjusted to the point where no steady-state
current in the bearing coils would be needed, thus reducing the power
consumption. There is some company (sunon) selling "maglev" fans, but as
far as I remember there is a point of contact in their design which in my
opinion means that it doesn't count.

Chris

 

[colin]

If your goal is rigidity I don't think you're going to get there with
magnetic bearings. You may be able to get the average measured error down
to zero, but if your shaft experiences any vibration it'll wobble, and
that'll show up as a wobble or wow in your position measurement in your
encoders.

thanks good point. but it might be able to run faster
this seems to be the main avantage, (and quieter),
they are used mainly in very high speed situations.
I thought magnetic bearings were suposed to be better than that.

If your real problem is the sensitivity of your angular measurement to
your shaft axis position, why don't you address that? A resolver,
properly built, should be significantly less sensitive to shaft run-out
issues than an encoder.

im not sure of the difference il have to lookit up,
but I see what you mean, maybe two opto encoders 1 on each side of the disc
will nullify the error.
I did think about that, but its more processing todo, id have to add another
micro to keep up
with 4 encoders or put the thing in a large fpga.

Should you be determined to stick with an encoder, why not take the same
shaft position measurements that you would have made for your magnetic
bearings, and just use them directly to correct the encoder output?

I'm not sure il think about that, its late in the morning atm.
it might also be friction cuasing torsional variations.

Colin =^.^=

 

[Rene Tschaggelar]

colin wrote:




Here's a magnetic bearing web page that I found interesting:
http://www.amsat.org/amsat/sats/phase3d/wheels/index.html

Magnetic bearings would be great for long-lasting fans, I think. The
position of the rotor could be adjusted to the point where no steady-state
current in the bearing coils would be needed, thus reducing the power
consumption. There is some company (sunon) selling "maglev" fans, but as
far as I remember there is a point of contact in their design which in my
opinion means that it doesn't count.

Chris

Physics requires a point of contact, unless you make
it vastly more complex without gaining much. Fans
create a thrust.

Rene

 

[Phil Hobbs]

Physics requires a point of contact, unless you make
it vastly more complex without gaining much. Fans
create a thrust.

Rene

You can make stable magnetic bearings with no contact by using eddy
current repulsion. An AC magnetic field in the neighbourhood of a
conductor causes currents to flow in the conductor that oppose the
applied field, resulting in repulsion. Using rotating fields, it's
quite possible to levitate and rotate a piece of metal with no contact
at all.

The problem with magnetic bearings in general, as others have pointed
out, is that they aren't very stiff, and if you want to do fancy things,
they tend to require closed loop control.

Air bearings are orders of magnitude stiffer, and also have no wear
(except during starting and stopping).

Cheers,

 

[colin]

You can make stable magnetic bearings with no contact by using eddy
current repulsion. An AC magnetic field in the neighbourhood of a
conductor causes currents to flow in the conductor that oppose the applied
field, resulting in repulsion. Using rotating fields, it's quite possible
to levitate and rotate a piece of metal with no contact at all.

The problem with magnetic bearings in general, as others have pointed out,
is that they aren't very stiff, and if you want to do fancy things, they
tend to require closed loop control.

Air bearings are orders of magnitude stiffer, and also have no wear
(except during starting and stopping).

Id looked at air bearings too, I could even have an air motor
however needing an air supply is a big drawback.
mostly noise.

the magnetic bearigns I was considering would of course have a control loop.
they are used in some very high speed compressors, turbines and centrifuges
I beleive.
things where I would think you need a lot of stiffness.

with a control loop you could make it almost infinitly stiff,
up to the point where it exceeds its limit,
although trying to make it stable and stiff and fast would be difficult.

the main advantage would probably be greatly increased speed capacity,
with no need for lubrication or noise and next to no friction.
im not sure if the mechanical error is from bearing play or shaft twisting
or chassis stiffness.
but more speed would reduce the effect of all these errors.

one thing I just thought of if i used closed loop magnetic bearings I could
then monitor the
error signal and see if this corelates to any thing in the signal.
Im not sure it would be worth just monitoring the radial play on its own for
this purpose.

incidently after 5 days of averaging the signal is still about 15ps, and the
noise has now gone down to <5ps

however 15ps is more than I was expecting, so I have to find a way to
'prove' that there is nothing else influencing the result at the same rate
as the signal wich is difficult as its period is 1.0027 days.
I was expecting to find anything from 0ps to about 3ps. although if you
subtract the noise 10ps is still realistic.

Colin =^.^=

 

[Phil Hobbs]

Id looked at air bearings too, I could even have an air motor
however needing an air supply is a big drawback.
mostly noise.

the magnetic bearigns I was considering would of course have a control loop.
they are used in some very high speed compressors, turbines and centrifuges
I beleive.
things where I would think you need a lot of stiffness.

It's only externally pressurized air bearings that need an air supply.

Since your shaft is obviously very well balanced and operating at a very
light load, you might be able to use a dynamically stabilized air
bearing (not sure of the right name)--as used in hard disc heads. You'd
still need a plain bearing for starting and stopping.

with a control loop you could make it almost infinitly stiff,
up to the point where it exceeds its limit,
although trying to make it stable and stiff and fast would be difficult.

Right. The bearing will get much softer at high frequency.

the main advantage would probably be greatly increased speed capacity,
with no need for lubrication or noise and next to no friction.
im not sure if the mechanical error is from bearing play or shaft twisting
or chassis stiffness.
but more speed would reduce the effect of all these errors.


one thing I just thought of if i used closed loop magnetic bearings I could
then monitor the
error signal and see if this corelates to any thing in the signal.
Im not sure it would be worth just monitoring the radial play on its own for
this purpose.

Interesting idea. If you're making that fancy a loop, you might be
able to make the shaft balance itself, using some nitinol, for instance.

incidently after 5 days of averaging the signal is still about 15ps, and the
noise has now gone down to <5ps

however 15ps is more than I was expecting, so I have to find a way to
'prove' that there is nothing else influencing the result at the same rate
as the signal wich is difficult as its period is 1.0027 days.
I was expecting to find anything from 0ps to about 3ps. although if you
subtract the noise 10ps is still realistic.

Colin =^.^=

Cheers,

Phil Hobbs