Ig, I'm having a thought that I can do a really bang-up job on the
precision of my lathe conversion with the most crappy ball screws on the
planet.

I have found some very affordable magnetic DRO scales and readers that
provide simple quadrature output (and a zero index) at TTL levels, and
can be had in both 2 micro-inch and 1 micro-inch resolution for very
little difference in price.

So, I'm thinking that instead of using the motor shaft encoders (which
are already there), I could use these scales to provide absolute
positioning accuracy down to .0002" in Z and .0001" in X, without much if
any changes in EMC's manner of looking at the data.

It seems to me that the "counts per revolution" datum could be used the
same with a linear scale as with a shaft encoder. Any errors in the
screws are simply going to be interpreted as speeding or slowing of the
CPR figure, rather than skewing the end position, so long as the motor
doesn't know or report what its rotational position is.

I think it might take a little code to get the motor to continue (or
stop) moving, unless I only use speed input on the drives, and not
positioning input.

What do you think of the idea of having absolute positioning information
in EMC, rather than the "dead reckoning" positioning provided by screws
of unknown accuracy? Do you know if anyone else has approached it this
way? (i.e. - taken a DRO, and used it for the positioning info?)

(Not just Ig... any other comments are welcome)

LLoyd

On 2011-09-10, Lloyd E. Sponenburgh lloydspinsidemindspring.com wrote:
Ig, I'm having a thought that I can do a really bang-up job on the
precision of my lathe conversion with the most crappy ball screws on the
planet.

I have found some very affordable magnetic DRO scales and readers that
provide simple quadrature output (and a zero index) at TTL levels, and
can be had in both 2 micro-inch and 1 micro-inch resolution for very
little difference in price.

So, I'm thinking that instead of using the motor shaft encoders (which
are already there), I could use these scales to provide absolute
positioning accuracy down to .0002" in Z and .0001" in X, without much if
any changes in EMC's manner of looking at the data.

It seems to me that the "counts per revolution" datum could be used the
same with a linear scale as with a shaft encoder. Any errors in the
screws are simply going to be interpreted as speeding or slowing of the
CPR figure, rather than skewing the end position, so long as the motor
doesn't know or report what its rotational position is.

I think it might take a little code to get the motor to continue (or
stop) moving, unless I only use speed input on the drives, and not
positioning input.

What do you think of the idea of having absolute positioning information
in EMC, rather than the "dead reckoning" positioning provided by screws
of unknown accuracy? Do you know if anyone else has approached it this
way? (i.e. - taken a DRO, and used it for the positioning info?)

(Not just Ig... any other comments are welcome)

Lloyd, this will work just fine. It is a known approach. If your DRO
is absolute, you can also avoid homing.

Igor

Ignoramus13208 fired this volley in
:

Lloyd, this will work just fine. It is a known approach. If your DRO
is absolute, you can also avoid homing.


The scales are not absolute, but do have a built-in home index, so switches
(or choppers) would not be required except to provide an over-limit panic
stop. The fact that the index is built into the scale would, I think,
further improve repeatablilty of homing. I've seen home switches go
"wonky" over time, and choppers can get dirty, and not repeat the same home
position twice in a row.

LLoyd

On Sat, 10 Sep 2011 14:46:27 -0500, "Lloyd E. Sponenburgh"
lloydspinsidemindspring.com wrote:

Ignoramus13208 fired this volley in
m:

Lloyd, this will work just fine. It is a known approach. If your DRO
is absolute, you can also avoid homing.


The scales are not absolute, but do have a built-in home index, so switches
(or choppers) would not be required except to provide an over-limit panic
stop. The fact that the index is built into the scale would, I think,
further improve repeatablilty of homing. I've seen home switches go
"wonky" over time, and choppers can get dirty, and not repeat the same home
position twice in a row.

LLoyd
Greetings LLoyd,
First of all, where are you getting the cheap magnetic scales?
Secondly, the biggest problem with using scales instead of a motor
shaft mounted encoder is backlash and resonance. As the control moves
the motor to correct the positional error if it has to turn the motor
shaft a lot before the scale moves then the motor can overshoot and
then this can lead to resonance problems. Inaccurate ballscrews aren't
a problem, just loose ballscrews.
Eric

fired this volley in
:

Inaccurate ballscrews aren't
a problem, just loose ballscrews.


I understand that. I'm not using loose ballscrews, I'm just resorting to
C7 precision screws rather than what I'd like to use -- can't afford the
good ones.

My ball nuts, end supports and thrust bearings all support essentially
zero backlash. I'm not using spring-loaded dual nuts, but thread pitch
skewed nuts that have an inherent preload.

The scales come from India, from Electronica Mechatronic Systems Pvt.
Ltd. I'm buying them on a "piggyback" order from DRO Pros in California.
(they get a cut, of course)

Their lead time is about three months, because they only import four
times a year, but they'll make up any scale you want from their stock
parts, and put it on the container. Mine are presently "stolen" from a
DRO using their parts, but I'll replace the DRO scales with the ones that
come in by December.

I mis-typed the precision of the scales. Mine are 1 micron (not micro-
inch) in X and 5 microns in Z. That makes a difference.

LLoyd

On 2011-09-10, Lloyd E. Sponenburgh lloydspinsidemindspring.com wrote:
fired this volley in
:

Inaccurate ballscrews aren't
a problem, just loose ballscrews.


I understand that. I'm not using loose ballscrews, I'm just resorting to
C7 precision screws rather than what I'd like to use -- can't afford the
good ones.

My ball nuts, end supports and thrust bearings all support essentially
zero backlash. I'm not using spring-loaded dual nuts, but thread pitch
skewed nuts that have an inherent preload.

The scales come from India, from Electronica Mechatronic Systems Pvt.
Ltd. I'm buying them on a "piggyback" order from DRO Pros in California.
(they get a cut, of course)

Their lead time is about three months, because they only import four
times a year, but they'll make up any scale you want from their stock
parts, and put it on the container. Mine are presently "stolen" from a
DRO using their parts, but I'll replace the DRO scales with the ones that
come in by December.

I mis-typed the precision of the scales. Mine are 1 micron (not micro-
inch) in X and 5 microns in Z. That makes a difference.

LLoyd


Quality of components aside, if everything works as intended, you
should be able to accomplish what you want.

i

"Lloyd E. Sponenburgh" lloydspinsidemindspring.com wrote:

Ig, I'm having a thought that I can do a really bang-up job on the
precision of my lathe conversion with the most crappy ball screws on the
planet.

I have found some very affordable magnetic DRO scales and readers that
provide simple quadrature output (and a zero index) at TTL levels, and
can be had in both 2 micro-inch and 1 micro-inch resolution for very
little difference in price.

The problem is if there is much backlash in the ballscrews, then you
will get servo "hunting" behavior, and it is pretty hard to get rid of
this. You can set a large deadband, but that is pretty counterproductive,
as it destabilizes the servo loop.

Jon

wrote in message ...

On Sat, 10 Sep 2011 14:46:27 -0500, "Lloyd E. Sponenburgh"
lloydspinsidemindspring.com wrote:

Ignoramus13208 fired this volley in
m:

Lloyd, this will work just fine. It is a known approach. If your DRO
is absolute, you can also avoid homing.


The scales are not absolute, but do have a built-in home index, so switches
(or choppers) would not be required except to provide an over-limit panic
stop. The fact that the index is built into the scale would, I think,
further improve repeatablilty of homing. I've seen home switches go
"wonky" over time, and choppers can get dirty, and not repeat the same home
position twice in a row.

LLoyd
Greetings LLoyd,
First of all, where are you getting the cheap magnetic scales?
Secondly, the biggest problem with using scales instead of a motor
shaft mounted encoder is backlash and resonance. As the control moves
the motor to correct the positional error if it has to turn the motor
shaft a lot before the scale moves then the motor can overshoot and
then this can lead to resonance problems. Inaccurate ballscrews aren't
a problem, just loose ballscrews.
Eric

I have a lathe and mill that works the way Lloyd is talking about. My
system works fine with mechanical slop in the system because it closes the
loop using tachometer feedback directly from the motor and closes the
position loop from the position encoder. My mill uses linear scales but the
lathe uses rotary encoders coupled to the ball screws. I would prefer the
lathe to have linear scales but at least with the encoders coupled to the
ball screws it takes out the motor/pulley/belt related errors, plus I get a
following error if a belt breaks hopefully instead of a destroyed part.

RogerN

Ignoramus13208 wrote:

On 2011-09-10, Lloyd E. Sponenburgh lloydspinsidemindspring.com wrote:


What do you think of the idea of having absolute positioning information
in EMC, rather than the "dead reckoning" positioning provided by screws
of unknown accuracy? Do you know if anyone else has approached it this
way? (i.e. - taken a DRO, and used it for the positioning info?)

(Not just Ig... any other comments are welcome)

Lloyd, this will work just fine. It is a known approach. If your DRO
is absolute, you can also avoid homing.
Well, if he has 1 uInch encoder resolution, then at 60 IPM he will have
a million encoder counts/second. That is fairly fast. Also, how will
he read this into EMC2? An existing quadrature encoder counter, or is
Lloyd going to build something custom to read the absolute data?

Jon

"Lloyd E. Sponenburgh" lloydspinsidemindspring.com wrote:

Ignoramus13208 fired this volley in
:

Lloyd, this will work just fine. It is a known approach. If your DRO
is absolute, you can also avoid homing.


The scales are not absolute, but do have a built-in home index, so
switches (or choppers) would not be required except to provide an
over-limit panic
stop. The fact that the index is built into the scale would, I think,
further improve repeatablilty of homing. I've seen home switches go
"wonky" over time, and choppers can get dirty, and not repeat the same
home position twice in a row.
For any highly-repeatable system, you usually have a home switch to get
close, and then search for the index pulse from the encoder to refine it
down to a single encoder count. This is how it is done on my Bridgeport,
using EMC2.

Jon

"Lloyd E. Sponenburgh" lloydspinsidemindspring.com wrote:



I understand that. I'm not using loose ballscrews, I'm just resorting to
C7 precision screws rather than what I'd like to use -- can't afford the
good ones.

OK, that is good. By the way, EMC2 also has leadscrew error correction.
So, you can create error compensation tables, and it will correct for
pitch variation in the screw.

The scales come from India, from Electronica Mechatronic Systems Pvt.
Ltd. I'm buying them on a "piggyback" order from DRO Pros in California.
(they get a cut, of course)

You really want to make sure these scales are DESIGNED for CNC motion
CONTROL applications, and not just DRO or static position reading use.
I just went through a mess with a CUI capacitive encoder that looked
great, but I couldn't tune the servo loop well. I finally discovered there
was a really nasty LAG in the encoder's response to acceleration. I found
it by putting a plain old optical encoder on the same motor shaft and
reading both into EMC at the same time. I worry that a 1 uInch scale
must be using interpolation also, and may suffer the same problem.

I mis-typed the precision of the scales. Mine are 1 micron (not micro-
inch) in X and 5 microns in Z. That makes a difference.
Ah, yes, that makes more sense, nobody makes much Imperial-measure stuff
anymore. 1 um is pretty good, about 39 uInch. 5 um is really just basic
DRO resolution, approximately .0002", and a bit rough for a high-accuracy
CNC machine.

Jon

Jon Elson fired this volley in
:

Also, how will
he read this into EMC2? An existing quadrature encoder counter, or is
Lloyd going to build something custom to read the absolute data?


That wouldn't matter much to me, Jon. I've built quadrature counters from
scratch before, including ones with edge multiplication.

LLoyd

Jon Elson fired this volley in
:

You really want to make sure these scales are DESIGNED for CNC motion
CONTROL applications, and not just DRO or static position reading use.
I just went through a mess with a CUI capacitive encoder that looked
great, but I couldn't tune the servo loop well. I finally discovered
there was a really nasty LAG in the encoder's response to
acceleration.

They guarantee these for output up to 2MHz (10m/s on the 5u scales, or
2m/s on the 1u ones). There doesn't appear to be any processing other
than signal conditioning going on in the reader heads.

Besides... I can do all this at NO risk. I have the DRO, I have encoders
on my motors, and I won't be destroying any of it to have a go at this
method. All I'd lose initially would be my time to try it, and the use
of the DRO display, but I'd gain that back with the EMC display.

LLoyd

On Sat, 10 Sep 2011 17:13:01 -0500, "Lloyd E. Sponenburgh"
lloydspinsidemindspring.com wrote:

Thanks Lloyd for the info.
Eric
fired this volley in
:

Inaccurate ballscrews aren't
a problem, just loose ballscrews.


I understand that. I'm not using loose ballscrews, I'm just resorting to
C7 precision screws rather than what I'd like to use -- can't afford the
good ones.

My ball nuts, end supports and thrust bearings all support essentially
zero backlash. I'm not using spring-loaded dual nuts, but thread pitch
skewed nuts that have an inherent preload.

The scales come from India, from Electronica Mechatronic Systems Pvt.
Ltd. I'm buying them on a "piggyback" order from DRO Pros in California.
(they get a cut, of course)

Their lead time is about three months, because they only import four
times a year, but they'll make up any scale you want from their stock
parts, and put it on the container. Mine are presently "stolen" from a
DRO using their parts, but I'll replace the DRO scales with the ones that
come in by December.

I mis-typed the precision of the scales. Mine are 1 micron (not micro-
inch) in X and 5 microns in Z. That makes a difference.

LLoyd

On Sat, 10 Sep 2011 23:06:08 -0500, Jon Elson wrote:
Ignoramus13208 wrote:
On 2011-09-10, Lloyd E. Sponenburgh wrote:

What do you think of the idea of having absolute positioning
information in EMC, rather than the "dead reckoning" positioning
provided by screws of unknown accuracy? Do you know if anyone else has
approached it this way? (i.e. - taken a DRO, and used it for the
positioning info?)
....
Lloyd, this will work just fine. It is a known approach. If your DRO is
absolute, you can also avoid homing.

Well, if he has 1 uInch encoder resolution, then at 60 IPM he will have
a million encoder counts/second. That is fairly fast. Also, how will
he read this into EMC2? An existing quadrature encoder counter, or is
Lloyd going to build something custom to read the absolute data?

In another post, Lloyd said resolution on one of the scales is
1 micron, rather than 1 microinch. At a million microns per meter,
1 micron is about 40 microinches. (.000039370")

--
jiw

On 09/11/2011 08:54 AM, Lloyd E. Sponenburgh wrote:
Jon fired this volley in
:

You really want to make sure these scales are DESIGNED for CNC motion
CONTROL applications, and not just DRO or static position reading use.
I just went through a mess with a CUI capacitive encoder that looked
great, but I couldn't tune the servo loop well. I finally discovered
there was a really nasty LAG in the encoder's response to
acceleration.

They guarantee these for output up to 2MHz (10m/s on the 5u scales, or
2m/s on the 1u ones). There doesn't appear to be any processing other
than signal conditioning going on in the reader heads.

I don't believe that. It would require there be magnetic stripes
recorded on the scales at a pitch of 1 um, which is pretty close!
Of course, they do this on computer disk drives, but the head rides
only a couple um above the platter there. The spacing on the read head
on a CNC measuring scale has to be bigger to allow for dust, swarf and
non-straightness. So, they really have to be interpolating the basic
scale resolution. How well they do that is up to the manufacturer.
I just wanted to throw out a warning that if the scales are made for
DRO use, only, then the latency between position changes and the
quadrature output may not be controlled. If people are using these for
closed-loop motion control, then this has probably already been dealt with.

I was pretty shocked that the CUI encoder had such a serious problem.
It is a commercial product from a major vendor, and sold specifically
for motion control use.
Besides... I can do all this at NO risk. I have the DRO, I have encoders
on my motors, and I won't be destroying any of it to have a go at this
method. All I'd lose initially would be my time to try it, and the use
of the DRO display, but I'd gain that back with the EMC display.
Well, that will work fine as a DRO, and you will just see how it works
as a closed-loop sensor. You can compare a rotary and linear encoder
on Halscope, like I did, to see if there are any anomalies.

Jon

Jon Elson fired this volley in
news
I don't believe that. It would require there be magnetic stripes
recorded on the scales at a pitch of 1 um, which is pretty close!

I don't think you understand how _multiple_ precisely-mounted read
elements in the reader head could be used to resolve dimensions much
smaller than the pole pitch -- so long as the pole pitch is uniform and
monotonic over its range.

I've done a lot of work in the past building quadrature pickups (from
scratch) and quadrature counters. If you have the precision available to
(say) mount four SETS of pickups at precisely 1/4 the "pitch" of the
"slot disk" (magnetic in this case), you get (inherently) four times the
pitch in precision. If you edge multiply the edge sensing by sinusoidal
pickup instead of square-wave choppers, you can reasonably multiply that
by 4 again without any significant loss of accuracy.

Now, take a pole pitch of .1mm (not all that fine). Multiply by (say) 10
sets of pickups at precisely measured distances between them, then do
edge multiplication x4, and you already get down to 25 microns.

I don't know how many sets of pickups there are, but 13 sets of pickups
gets it down below 1 micron. That's pretty "do-able" with current
technology.

LLoyd