开云体育

From: Jon Elson <jmelson@...>
Subject: Re: Servo Amp designs

Ohh, there's the killer, right there. Just like all the other darn
beautiful one-axis chips (Nat Semi LM628/629, HP HCTL1200
etc.) when you need coordinated motion on 2 - n axes, it all falls
apart. Especially for contouring or engraving, where you're following
splines with a change in angle every few thousandths of an inch.
That's the reason why everyone seems to come up with some way
of using one processor to compute the trajectory for ALL the axes.

Yep, I've been forced to use single axis controllers to perform
interpolated moves. It's not the best way of doing it to say the least.
A multi-axis approach, though more expensive, is probably required.

From: Tom Kulaga <tkulaga@...>
Subject: Re: Digest Number 19

If you plan to use steppers, well, you're running open-loop already,
but I guess you could use the linear encoders as feedback to correct
for missed steps. This assumes that you'll be writing a custom
program to run the steppers, unless there's software out there to
take advantage of the encoder feedback. I haven't seen any, but then
again, I haven't looked. :)

Some of the commercial stepper cards do this, they move, check the
final position, and make corrections as specified by various
parameters. Not great, but workable. Especially since you then
know you had slippage or a stall and can correct for it or at
least stop the process.

From: Jon Elson <jmelson@...>
There is an Atlas/Craftsman list, and the metalworking web site has a fAQ,
already. It probably has little specifically on CAD, CAM and EDM.
It does have some pointers and comments about DROs.

I'll have to check that one out!

Jim Eckman

Tom,

I am definitely interested. I noticed that US Digital also offers
linear encoders for $2 per inch, which sounds reasonable. Of course
the electronic "readers" are extra, but the cost should be about the
same as for as a rotary (I presume!). Wouldn't this help address
backlash? At least you should be able to tell where you actually are.
I'm not sure how the slop effect when changing directions - not likely
to be good. But wouldn't this be a problem even with backlash
compensation?

Steve

At 10:02 PM 5/21/99 -0700, you wrote:

From: Tom Kulaga <tkulaga@...>

I guess you've all seen the posts about the DRO card I built. I've
got an update to the software, adding metric support, cleaning up a
few things I didn't like, and changing the 4th axis to display in
degrees instead of linear units. I haven't posted it to the webpage
yet because I wanted to add backlash compensation first. That's the
problem.

The big drawback of backlash compensation is that you are forced to
make sure that all of the slop is cranked to one side of the travel
before you can zero that axis. If you don't, the computer doesn't
know where you are inside the backlash zone, and you can't trust the
numbers on the display.

My question is this: Do those of you who want to use rotary encoders
have a problem with this? Keep in mind, you won't be able to zero
the axis until you are certain that you've moved all the way to the
left or the right of the slop. I can think of more than one occasion
where this would be *really* annoying, but usually it would just be a
matter of getting in the habit of watching what you're doing a bit
more so than normal.

So, did I get that across clear enough? Is it worth it to add this
feature? I don't want to spend a whole bunch of time working on a
feature that nobody will want to use, especially since I've got a new
project I want to start.

Speaking of that new project, I've been following the comments about
the EMC software (gotta love the free stuff ;) with interest, and got
to thinking about how I'd go about trying it out here at home. I was
thinking that buying a motion control card is pretty pricey, and that
my DRO card makes a great jumping-off point for bigger and better
things. With a little extra work, and probably an extra 200 bucks, I
could come up with a 3-axis servo controller card design. It's a
little more involved, but using PWM for the output signal cleans it up
nicely, bringing it into the 'definately doable' range, even for
beginners.

I think that this group could easily write plans to build the card,
the servo amps, and the guidance you'd need to assemble the lot into
a really good webpage. I can handle the card OK, but I need to know
how the card interfaces with the EMC stuff, or if I could provide all
the register-level programming info to somebody who could write a
driver or whatever is needed to get things going. I don't yet have
Linux here at home, so I can't really experiment much. If there's
interest in this, I'll order Redhat next month.

What do you say, guys? Sound like fun?

-Tom Kulaga



------------------------------------------------------------------------
With more than 17 million e-mails exchanged daily...

...ONElist is THE place where the world talks!
------------------------------------------------------------------------
welcome to CAD_CAM_EDM_DRO@..., an unmodulated list for the

discussion of shop built systems in the above catagories.

I would be willing to help build theprinted circuit card. But take a look at
the PIC chip from JR Kerr
They have a PIC-Servo chip that has PID filter and all in a $30 chip.
The web site is www.jrkerr.com The pic servo may decrease the effort
substanially. I have tried a single axis version of the chip and it is
fast. The only problem I ran into was that he sells a PIC-Step chip that
allows use of the parallel port and existing step and direction software.
The servo runs very slowly on a 486-33SX. The best I could get was about 200
RPM with a 200 line encoder. A 500 line encoder only yielded 100 RPM
It may have been due to the 486-33SX but by using the ISA bus it would run
and 2000 RPM. When I ran the pic servo using the RS232 port I was able to
get well over 5000 RPMs.
Dan

My mill drill did at least 66 IPM . But I have ballscrews on the x and Y
axis and was using my 5 A drivers for the motor power . The motors were 300
ounce inch motors.
I have solved the quill backlash problem and now have .0019 backlash in the
quill.
Dan

I will have the parts done next week.
Dan

At 10:02 PM 5/21/99 -0700, you wrote:

From: Tom Kulaga <tkulaga@...>

Speaking of that new project, I've been following the comments about
the EMC software (gotta love the free stuff ;) with interest, and got
to thinking about how I'd go about trying it out here at home. I was
thinking that buying a motion control card is pretty pricey, and that
my DRO card makes a great jumping-off point for bigger and better
things. With a little extra work, and probably an extra 200 bucks, I
could come up with a 3-axis servo controller card design. It's a
little more involved, but using PWM for the output signal cleans it up
nicely, bringing it into the 'definately doable' range, even for
beginners.

I think that this group could easily write plans to build the card,
the servo amps, and the guidance you'd need to assemble the lot into
a really good webpage.

If there's

interest in this, I'll order Redhat next month.

What do you say, guys? Sound like fun?

-Tom Kulaga

Tom,
I would be interested. The high price of the Servo-to-Go card is one thing
that killed my interest in the servos. If you can do it for $250-$300, it
would definitely kick it in gear for me.

Thanks,
Dan

Speaking of that new project, I've been following the comments about
the EMC software (gotta love the free stuff ;) with interest, and got
to thinking about how I'd go about trying it out here at home. I was
thinking that buying a motion control card is pretty pricey, and that
my DRO card makes a great jumping-off point for bigger and better
things. With a little extra work, and probably an extra 200 bucks, I
could come up with a 3-axis servo controller card design. It's a
little more involved, but using PWM for the output signal cleans it up
nicely, bringing it into the 'definately doable' range, even for
beginners.

I think that this group could easily write plans to build the card,
the servo amps, and the guidance you'd need to assemble the lot into
a really good webpage. I can handle the card OK, but I need to know
how the card interfaces with the EMC stuff, or if I could provide all
the register-level programming info to somebody who could write a
driver or whatever is needed to get things going. I don't yet have
Linux here at home, so I can't really experiment much. If there's
interest in this, I'll order Redhat next month.

What do you say, guys? Sound like fun?

Yes, it does sound very fun to me and I do want to keep in touch with all the people that will be working on this. I am very much in favor of your ideas there Tom, and thanks for the reply on Dan's card.

Regards,
Don Hughes
pencad@...

------------------------------------------------------------------------
With more than 17 million e-mails exchanged daily...

...ONElist is THE place where the world talks!
------------------------------------------------------------------------
welcome to CAD_CAM_EDM_DRO@..., an unmodulated list for the discussion of shop built systems in the above catagories.

i saw a company with a rt-linux release cdrom. check out
www.zentropix.com

here's a paragraph from their product page:

The Zentropix CD-ROM for RTLinux real-time extensions contains a
complete copy of the Red Hat version 5.2 Disk #1, but with a modified
kernel to support hard real-time functions. It supports all the
functions of the unmodified kernel, and is indistinguishable to the
non-RT user. However, to any user looking for the benefits of a
deterministic, hard real-time operation, this provides a user-friendly
installation for the New Mexico Tech kernel modification (currently
2.0.36).

their price is $35 plus shipping.

this implies an easier way to get running for a first time Linux user.

-ron

Dan Mauch wrote:

From: "Dan Mauch" <dmauch@...>

I can see that in time we will be answering the same questions over and over
and over again. I think there need to be a FAQ set up so these repetitive
questions can be answered. That way the focus can remain on developing new
hardware for the home shops likes Tom's DRO and Jon's .
Unfortuneately, I don't have the time to set one up. Any volunteers.

I'd be glad to write up some stuff on CNC, especially some of the details of
G-code and servo systems. I also have a bunch of stuff I've written over
the past few years on Atlas lathes and Bridgeport mills. Most of that is
replies to people's question, but I could search out some of the more general
pieces and put them into some order.

There is an Atlas/Craftsman list, and the metalworking web site has a fAQ,
already. It probably has little specifically on CAD, CAM and EDM.
It does have some pointers and comments about DROs.

Jon

Jon Elson wrote:

Now, if you mean just to use the linear encoders, that's another story.

Jon

Using the "glass scales" was actually what I was thinking of (mitutoyo). If
anyone has info on the outputs of the scales, that would be helpful, but
if not, I'll digout the scope, etc. and find out what comes out. (In one
of my past lives I was paid to do reverse engineering)

The Digital Readout Unit itself has a cutout on the back panel for a DB25
connector, and is labelled aux out (or something like that), but I haven't ever
opened it up, to see what is missing (or empty holes on the PCB, not stuffed).

Regards,
George Potter

<PTENGIN@...> wrote:

In looking over information from Mitutoyo, they offer BCD and several
other output cards for the KC and other DRO systems. I have one of these
(KC) systems and was suspecting the output could be used to inteface to a
CNC. Also, in the MSC catalog, Mitutoyo used to list a linear spar to digital
output converter box. Peter

One big problem using the output of the linear encoders of the DRO to
control servos is that *any* backlash between the drive motor and the
workpiece can cause some pretty bad instability (for controls
newbies, the instability I'm talking about is a tendancy for the
motor to oscillate around the target position, buzzing like crazy in
the best case, or slamming back and forth in the worst). There are
ways around it, but none of them are cheap, assuming you are faced
with the normal HSM budget. And no, even zero-backlash ballscrews
won't be enough to eliminate the backlash unless you're running VERY
light loads at low speeds. I had once guessed otherwise, but found
out fast that I was wrong.

If you plan to use steppers, well, you're running open-loop already,
but I guess you could use the linear encoders as feedback to correct
for missed steps. This assumes that you'll be writing a custom
program to run the steppers, unless there's software out there to
take advantage of the encoder feedback. I haven't seen any, but then
again, I haven't looked. :)

-Tom Kulaga

I guess you've all seen the posts about the DRO card I built. I've
got an update to the software, adding metric support, cleaning up a
few things I didn't like, and changing the 4th axis to display in
degrees instead of linear units. I haven't posted it to the webpage
yet because I wanted to add backlash compensation first. That's the
problem.

The big drawback of backlash compensation is that you are forced to
make sure that all of the slop is cranked to one side of the travel
before you can zero that axis. If you don't, the computer doesn't
know where you are inside the backlash zone, and you can't trust the
numbers on the display.

My question is this: Do those of you who want to use rotary encoders
have a problem with this? Keep in mind, you won't be able to zero
the axis until you are certain that you've moved all the way to the
left or the right of the slop. I can think of more than one occasion
where this would be *really* annoying, but usually it would just be a
matter of getting in the habit of watching what you're doing a bit
more so than normal.

So, did I get that across clear enough? Is it worth it to add this
feature? I don't want to spend a whole bunch of time working on a
feature that nobody will want to use, especially since I've got a new
project I want to start.

Speaking of that new project, I've been following the comments about
the EMC software (gotta love the free stuff ;) with interest, and got
to thinking about how I'd go about trying it out here at home. I was
thinking that buying a motion control card is pretty pricey, and that
my DRO card makes a great jumping-off point for bigger and better
things. With a little extra work, and probably an extra 200 bucks, I
could come up with a 3-axis servo controller card design. It's a
little more involved, but using PWM for the output signal cleans it up
nicely, bringing it into the 'definately doable' range, even for
beginners.

I think that this group could easily write plans to build the card,
the servo amps, and the guidance you'd need to assemble the lot into
a really good webpage. I can handle the card OK, but I need to know
how the card interfaces with the EMC stuff, or if I could provide all
the register-level programming info to somebody who could write a
driver or whatever is needed to get things going. I don't yet have
Linux here at home, so I can't really experiment much. If there's
interest in this, I'll order Redhat next month.

What do you say, guys? Sound like fun?

-Tom Kulaga

In a message dated 5/21/99 7:00:18 PM Hawaiian Standard Time,
tkulaga@... writes:

<< unless there's software out there to
take advantage of the encoder feedback. I haven't seen any, but then
again, I haven't looked. :)

-Tom Kulaga >>

Tom,
Thanks for the reply. I was looking at the Microkinetics Package. It does
have a provision for encoder feed back. I realize there are two things we are
looking at. One would be lost steps, the other is location error feed back.
If I simply put encoders on my Acme lead screws, I could only keep track of
steps but not error. This is where I was wondering if adding spars would be a
cheaper and/or easier way of accurately positioning the work rather than
ballscrews. I have come to the conclusion that I am barking up the wrong
tree and need to get the ballscrews first. Peter

In a message dated 5/21/99 3:36:21 AM Hawaiian Standard Time,
dmauch@... writes:

<< I used 300 oz in stepper on my mill drill with a 2/1 ratio timing belts and
pulleys. >>
Dan,
Just how fast does this setup go? I tried 300's on a FB-2 direct drive
with poor results. I figured it was a smaller machine. I will be installing
the 2:1 reduction in a week or so... I'm fighting the backlash also. The z
axis is worst on climbing to home. I'm gonna try nitrogen springs from a
hatch back car. Peter

Hi to all,

Does any body have Robert Langlois email address? Or know
if he is on vacation at the moment? Sent a message to him at the
address he had about 18 months ago and haven't seen a reply in over
a week. My current projects are an EDM machine and trying to talk a
friend into converting his 6" atlas lathe to cnc. Would have more
projects but I'm job hunting and have to limit expenses. Also why
don't people on this list only quote the part of someone else email
that has to do with there question/answer instead of the entire message?
I see a fair amount of messages with a few new lines of information and
50 lines of old info not even relivent to the new stuff?

Mark Holycross

P.S. Had a great time at NAMES and was lucky enough to meet Dan & partner
(sorry I forgot the other gents name) and Robert at there cnc&edm seminars.

From: Jon Elson <jmelson@...>
paul@... wrote:

Using the PIC for interface as well, it gets driven through a
standard RS232 port using various commands (absolute or relative
moves, velocity, etc). The code incorporates a tunable PID
algorithm, velocity ramps, etc. Using a 16Mhz clock, it can
run a PID calculation every .25 milisecond.

Ohh, there's the killer, right there. Just like all the other darn
beautiful one-axis chips (Nat Semi LM628/629, HP HCTL1200
etc.) when you need coordinated motion on 2 - n axes, it all falls
apart. Especially for contouring or engraving, where you're following
splines with a change in angle every few thousandths of an inch.
That's the reason why everyone seems to come up with some way
of using one processor to compute the trajectory for ALL the axes.

Good point; the appnote is really talking about a 1 axis solution.

--
Paul Amaranth | Rochester MI, USA
Aurora Group, Inc. | Software Development
paul@... | Unix / C / Tcl-Tk

paul@... wrote:

I was poking through my reference stack the other night and came
across an interesting digital servo amp design using a PIC microcontroller.
There are a couple of interesting things about it.

It uses a digital encoded on the motor for both position and velocity
information. In a rather slick little trick, it takes the quadrature
output of the encoded and uses a PLD to turn it into a stream of
up or down pulses (with selectable 1x or 4x resolution). These
go directly to the PIC's internal counters.

The PIC uses its builtin PWM hardware to generate the motor drive
signals which, in the app note, go to an integrated H bridge
driver.

Using the PIC for interface as well, it gets driven through a
standard RS232 port using various commands (absolute or relative
moves, velocity, etc). The code incorporates a tunable PID
algorithm, velocity ramps, etc. Using a 16Mhz clock, it can
run a PID calculation every .25 milisecond.

Ohh, there's the killer, right there. Just like all the other darn
beautiful one-axis chips (Nat Semi LM628/629, HP HCTL1200
etc.) when you need coordinated motion on 2 - n axes, it all falls
apart. Especially for contouring or engraving, where you're following
splines with a change in angle every few thousandths of an inch.
That's the reason why everyone seems to come up with some way
of using one processor to compute the trajectory for ALL the axes.

The hardware count is pretty low, maybe 8-10 ICs and a handful of
discrete parts, although, as a trade off, the code is fairly
involved. With a little additional tweaking, it might be
the basis of a very useful servo amp (use a different H-Bridge,
make the PID parameters loadable from the interface, etc).

Then, of course, there is the problem of integrating the interface
with a G code interpreter.

I'm rather interested in building one of these and seeing how
well a run of the mill DC motor with an encoder would function.
I have a number of 12V motors that have reasonable speed and
torque characteristics for small CNC work, although I wouldn't
put them on a Bridgeport.

Given the state of my project queue, that ain't gonna happen
this year.

The design was in the PIC Application Handbook, developed around
'93 if anyone is interested.

If anyone runs across any other public servo amp designs, I'd
like to hear about them.

Well, my PWM servo amps aren't exactly public domain, as I want to
retain the rights to make these as a commercial product in the future,
there are about 12 people who have bought my 'micro kits' for them.
That is a PCB and a few components that bould be hard to get in
small quantities, plus all documentation on them. (They do need
DC tachs, and take a +- 10 V velocity command from a DAC on
the CNC control.) It might be a neat project to make an small
board that converts quadrature encoder signals to emulate an analog
DC tach. I think it could be made with about 3 chips and half a
dozen discrete parts. But, it would be hard to make it as clean
as the DC tachs can. I can have the machine move down to a
feed rate of .01"/Minute or so before it starts to exhibit stick/slip
friction. That is about 4 encoder counts per second on my particular
setup, so I don't know how smoothly the system would respond
down there. But, that is not a real common situation, anyway.

Jon

Jon,

It seems like I could rig a tach to the back end of my servo motors if I
wanted , just as long as I got some kind of tach output eh? Then I could
use your servo amps?

How much would a set of 4 servo boards go for? Or the kits? I am
interested.

I'm trying to build my own small machine and need a source for smaller
ballscrews, about 12" travel zero backlash, 3 of them. Any ideas. I'm
trying to do it on the cheap so surplus would be ok. I even have a 36" or
so travel super precision ground ball screw and nut I could trade....

Michael

At 02:27 PM 5/21/99 -0500, you wrote:

From: Jon Elson <jmelson@...>



paul@... wrote:

I was poking through my reference stack the other night and came
across an interesting digital servo amp design using a PIC microcontroller.
There are a couple of interesting things about it.

It uses a digital encoded on the motor for both position and velocity
information. In a rather slick little trick, it takes the quadrature
output of the encoded and uses a PLD to turn it into a stream of
up or down pulses (with selectable 1x or 4x resolution). These
go directly to the PIC's internal counters.

The PIC uses its builtin PWM hardware to generate the motor drive
signals which, in the app note, go to an integrated H bridge
driver.

Using the PIC for interface as well, it gets driven through a
standard RS232 port using various commands (absolute or relative
moves, velocity, etc). The code incorporates a tunable PID
algorithm, velocity ramps, etc. Using a 16Mhz clock, it can
run a PID calculation every .25 milisecond.

Ohh, there's the killer, right there. Just like all the other darn
beautiful one-axis chips (Nat Semi LM628/629, HP HCTL1200
etc.) when you need coordinated motion on 2 - n axes, it all falls
apart. Especially for contouring or engraving, where you're following
splines with a change in angle every few thousandths of an inch.
That's the reason why everyone seems to come up with some way
of using one processor to compute the trajectory for ALL the axes.

The hardware count is pretty low, maybe 8-10 ICs and a handful of
discrete parts, although, as a trade off, the code is fairly
involved. With a little additional tweaking, it might be
the basis of a very useful servo amp (use a different H-Bridge,
make the PID parameters loadable from the interface, etc).

Then, of course, there is the problem of integrating the interface
with a G code interpreter.

I'm rather interested in building one of these and seeing how
well a run of the mill DC motor with an encoder would function.
I have a number of 12V motors that have reasonable speed and
torque characteristics for small CNC work, although I wouldn't
put them on a Bridgeport.

Given the state of my project queue, that ain't gonna happen
this year.

The design was in the PIC Application Handbook, developed around
'93 if anyone is interested.

If anyone runs across any other public servo amp designs, I'd
like to hear about them.

Well, my PWM servo amps aren't exactly public domain, as I want to
retain the rights to make these as a commercial product in the future,
there are about 12 people who have bought my 'micro kits' for them.
That is a PCB and a few components that bould be hard to get in
small quantities, plus all documentation on them. (They do need
DC tachs, and take a +- 10 V velocity command from a DAC on
the CNC control.) It might be a neat project to make an small
board that converts quadrature encoder signals to emulate an analog
DC tach. I think it could be made with about 3 chips and half a
dozen discrete parts. But, it would be hard to make it as clean
as the DC tachs can. I can have the machine move down to a
feed rate of .01"/Minute or so before it starts to exhibit stick/slip
friction. That is about 4 encoder counts per second on my particular
setup, so I don't know how smoothly the system would respond
down there. But, that is not a real common situation, anyway.

Jon


------------------------------------------------------------------------
How many communities do you think join ONElist each day?

More than 1,000! Create yours now!
------------------------------------------------------------------------
welcome to CAD_CAM_EDM_DRO@..., an unmodulated list for the

discussion of shop built systems in the above catagories.

--
michael leonard tilse - draemr@...
tujunga.california.usa.terra.sol.milkyway

John:

On the demo that I downloaded, it made mention of a no-save item as well
as another cut point of a 1000 lines of code. But not that I've seen any
cut-off date.

Regards
Don



John Stevenson wrote:

Don Hughes wrote:

The question I have now is, does anyone know the process for nesting
multiple parts into BobCAD? I tried the "copy", but that did not work.
Would I have to draw the multiple parts in AutoCAD and then export that
drawing with the cutting table dimensions already in place. I will have
a cutting space of 24" x 48".

First, you select inside area, or select all, or in some other way select
what you want to copy. Then, you do a change, translate, and enter
the amount you want the copy to be offset from the original, and set
the incremental and make copy buttons, and you'll get an additional
copy of the part, moved over that much. Just keep doing that to
make as many copies as you need.

Jon