Re: acceleration profille
Posted by
mariss92705
on 2002-07-27 19:52:51 UTC
Hi,
There is no "best" acceleration profile.
Linear acceleration is "best" because of simplicity. You just add a
constant to velocity per unit of time to get linear acceleration. Its
drawbacks are less than optimum time to speed and a large 2nd
derivative, called the "jerk factor".
Pretend you are in a car decelerating to a stop at a red light. You
are braced against the g-force of deceleration and are sitting still.
The moment the car comes to a stop, deceleration abruptly disappears
and you are still braced against a force that just vanished. Until
you adjust yourself, your head and upper body will slam backwards.
This is the jerk factor. A moving piece of machinery will resonate
or "ring" under the same conditions.
S-shaped acceleration curves take care of the "jerk" factor the same
way you take care of it in the car in the above example. Just before
coming to a complete stop you ease off the break to make the stop
more comfortable. At the start of a move, acceleration is very gentle
but increases with time to a maximum level. It then decreases again,
gently tailing off to zero when the target velocity is reached. The
drawbacks are complexity and an even slower time to target speed.
The best time to speed uses an acceleration curve that is the first
integral of the motor's speed-torque curve. For a DC servo motor it
would be parabolic, for a step motor it would be a higher order
parabolic. The drawbacks are complexity and a significant jerk factor.
Mariss
There is no "best" acceleration profile.
Linear acceleration is "best" because of simplicity. You just add a
constant to velocity per unit of time to get linear acceleration. Its
drawbacks are less than optimum time to speed and a large 2nd
derivative, called the "jerk factor".
Pretend you are in a car decelerating to a stop at a red light. You
are braced against the g-force of deceleration and are sitting still.
The moment the car comes to a stop, deceleration abruptly disappears
and you are still braced against a force that just vanished. Until
you adjust yourself, your head and upper body will slam backwards.
This is the jerk factor. A moving piece of machinery will resonate
or "ring" under the same conditions.
S-shaped acceleration curves take care of the "jerk" factor the same
way you take care of it in the car in the above example. Just before
coming to a complete stop you ease off the break to make the stop
more comfortable. At the start of a move, acceleration is very gentle
but increases with time to a maximum level. It then decreases again,
gently tailing off to zero when the target velocity is reached. The
drawbacks are complexity and an even slower time to target speed.
The best time to speed uses an acceleration curve that is the first
integral of the motor's speed-torque curve. For a DC servo motor it
would be parabolic, for a step motor it would be a higher order
parabolic. The drawbacks are complexity and a significant jerk factor.
Mariss
--- In CAD_CAM_EDM_DRO@y..., Alan Marconett KM6VV <KM6VV@a...> wrote:
> Hi Michael,
>
> A linear acceleration or deceleration might be easier to do, but
not the
> best. You might take a look at a manual for the CY545 stepper
> controller from Cybernetic Microsystems.
>
> http://controlchips.com
>
> Look at the rate tables, and the plots of
acceleration/deceleration.
> There's also some useful information on steppers in the manual.
>
> I use a similar table in my controller program to do
> acceleration/deceleration. An acceleration term determines how
fast the
> controller program runs up/down the table.
>
> Don't know what you mean by "skipping a step" at almost full
speed. The
> time between steps simply changes.
>
> Alan KM6VV
>
>
> Michael Holm wrote:
> >
> > I also made a stepper controlling software. This one runs under
windows, and
> > does linar acceleration.
> >
> > I did some thinking on that, and was about to make a user-editable
> > acceleration curve, because I was in doubt if it should be
linar.But if you
> > think hard about it, it makes no sense to make it curved curve,
as that
> > would mean that you'd have to skip just one step at some point,
when running
> > almost full speed, and that have to be even harder then starting
at full
> > speed.
> >
> > So it has to be linar interpolated.
> >
> > Draw a curved acceleration curve on a piece of squared paper
(using only
> > full squares, representing one step) and you'll understand what I
mean.
> >
> > --Michael
Discussion Thread
keongsan
2002-07-26 02:01:35 UTC
acceleration profille
Dan Mauch
2002-07-26 11:19:31 UTC
RE: [CAD_CAM_EDM_DRO] acceleration profille
Dave Hylands
2002-07-26 11:31:42 UTC
RE: [CAD_CAM_EDM_DRO] acceleration profille
cadcambee
2002-07-26 17:21:48 UTC
Re: acceleration profille
keongsan
2002-07-27 05:04:39 UTC
Re: acceleration profille (gained steps?)
jmkasunich
2002-07-27 12:26:06 UTC
Re: acceleration profille
Graham Hollis
2002-07-27 15:05:11 UTC
Re: [CAD_CAM_EDM_DRO] Re: acceleration profille
Michael Holm
2002-07-27 16:49:00 UTC
Re: [CAD_CAM_EDM_DRO] acceleration profille
jmkasunich
2002-07-27 17:02:37 UTC
Re: acceleration profille
Alan Marconett KM6VV
2002-07-27 17:30:05 UTC
Re: [CAD_CAM_EDM_DRO] acceleration profille
cadcambee
2002-07-27 18:08:52 UTC
Re: acceleration profille
mariss92705
2002-07-27 19:52:51 UTC
Re: acceleration profille
Alan Marconett KM6VV
2002-07-28 11:26:47 UTC
Re: acceleration profille
keongsan
2002-07-28 16:57:15 UTC
Re: acceleration profille (gained steps?)
keongsan
2002-07-28 17:01:16 UTC
Re: acceleration profille (gained steps?)
Steve Blackmore
2002-07-28 17:34:56 UTC
Re: [CAD_CAM_EDM_DRO] Re: acceleration profille
Dave Hylands
2002-07-28 18:00:37 UTC
RE: [CAD_CAM_EDM_DRO] Re: acceleration profille (gained steps?)
mariss92705
2002-07-28 18:20:29 UTC
Re: acceleration profille (gained steps?)
Alan Marconett KM6VV
2002-07-28 19:57:30 UTC
Re: acceleration profille
MIKEC@W...
2002-07-28 21:41:19 UTC
Re: [CAD_CAM_EDM_DRO] Re: acceleration profille
MIKEC@W...
2002-07-28 21:44:17 UTC
Re: [CAD_CAM_EDM_DRO] acceleration profille
chapmani
2002-07-29 02:11:59 UTC
Re: acceleration profille
eforum3001
2002-07-29 04:04:01 UTC
Re: acceleration profille (gained steps?)