开云体育

This was intended in a joking fashion. No serious person would suggest that
any reader of this list is anything other than hardworking and clever.
From griping comes invention, as it is an intermediate state after
necessity.

In spite of the request to forgo handwaving, that's where I'll begin:
The light source is the easiest part, but there are some subtlies: the
light should be nearly collimated. The simplest way to achieve this is to
use a LED with a narrow beam angle. The standard T1 3/4 lamps are 5 mm in
diameter and have a beam a bit smaller than that. There are some larger
diameters which could be handy.

Detectors are another issue. You can use 2 or 4 detectors. I prefer the 4
detector approach, since the electronics then don't have to be as stable.
There are some 4 detector packages, for example the UDT SPOT-4DMI and the
FIL-S4DG, these have detector sizes of 0.5 mm square and 1.3 mm square,
respectively.
It is desirable to have the detectors cover as little area as possible, so
that the system will be less sensitive to angular misalignment.

The enterprising builder might remove the quadrant detector from a CD
player. Personally I don't like fussing with surface mount packages.

Now on to getting patterns printed on plastic. Forgive me if this is too
elementary, but there may be some out there who don't know this trick. The
trick is to use the Linotronic (type) film printer at a printing service.
These print on mylar film very high contrast stable patterns. The machines
have resolutions from between 1200 and 3600 dpi.
They are used by Mac zealots, and only understand file formats that are
popular in the graphic arts world. My advise is to not try to get them to
understand any of your file formats, rather to find out what Mac's like and
use that. The programs they typically use are Pagemaker and Photoshot. So
if you can make a file in one of those, you're all set. They also read eps
files, but I've found that these give poor results- maybe you'll have better
results than me. AutoCAD can output eps files, but the linotronic printers
don't interpret them accurately.
I output files in dxf format and then make a Pagemaker document. This works
quite well.
You might want to consider how the machines native dot spacing can alias at
the spacing you want to use. Tilting the pattern at an angle to the machine
axes can reduce the effect of the aliasing.

If a few people are interested in making these, let me know and I'll work up
a parts list and a graphics file.

Mount the LED so it illuminates the detectors. A pattern drawn with line
patterns stepped in phase 90 is aligned to the detector. Op amps are used
to amplify the signal from the detector, using a transimpedance
configuration. The signals from the 0 and 180 are differenced with another
op amp, as are the 90 and 270 signals. The outputs from the differential
amplifiers are fed to triggers that go to 1 when the output is positive and
0 whent the output is negative.

The people on this list can surely think of ways to use the precise patterns
that the linotronic machines can generate. You can make scales, half-tone
patterns, and my favorite, the business card with microscopic features on
it.

Finally, let me add my voice to those who think that a CD with RT Linux and
EMC on it made as simple as possible to install is a good idea. Like, put
in the disk, get a instant machine controller. Is this possible?

Elliot Burke

"Elliot Burke" <elliot@...> writes:

Now on to getting patterns printed on plastic. Forgive me if this is too
elementary, but there may be some out there who don't know this trick. The
trick is to use the Linotronic (type) film printer at a printing service.
These print on mylar film very high contrast stable patterns. The machines
have resolutions from between 1200 and 3600 dpi.

A good idea, subject to the proviso that film recorders are not perfectly
linear, and that high resolution alone is no guarantee of linearity and
absolute accuracy. In my experience most low-end service bureau shop owners
don't know the specifications of their machines, just whether or not it is
adequate for their application. For precision work I recommend you find a
service bureau that does specializes in quality color-separation work and
image it at the highest resolution available, preferably 3600 dpi.

They are used by Mac zealots, and only understand file formats that are
popular in the graphic arts world.

They all understand PostScript and will accept a PostScript text file for
direct download. For a project like this it would be preferable to avoid
page layout programs and code the scale directly in PostScript to avoid
pixel rounding errors, although if you run it at 3600 dpi the rounding error
may not be significant enough to matter. If line width is critical, note
that rounding errors can give rise to lines of different width as well as
+/- 1 pixel positioning errors.

Finally, let me add my voice to those who think that a CD with RT Linux and
EMC on it made as simple as possible to install is a good idea. Like, put
in the disk, get a instant machine controller. Is this possible?

Since the installation of Linux is very machine dependent, i.e., the set of
drivers installed on my machine may differ from the set installed on another
machine of different hardware makeup, I suspect it won't be possible, or at
least simple, in the near term.

Phil Plumbo
St. Paul, MN

Elliot Burke wrote:

In spite of the request to forgo handwaving, that's where I'll begin:
The light source is the easiest part, but there are some subtlies: the
light should be nearly collimated. The simplest way to achieve this is to
use a LED with a narrow beam angle. The standard T1 3/4 lamps are 5 mm in
diameter and have a beam a bit smaller than that. There are some larger
diameters which could be handy.

Visible red lasers are now available for about $9. By running them below
normal rated power, they will last a very long time. They are VASTLY
more collimated than LEDs.

Detectors are another issue. You can use 2 or 4 detectors. I prefer the 4
detector approach, since the electronics then don't have to be as stable.
There are some 4 detector packages, for example the UDT SPOT-4DMI and the
FIL-S4DG, these have detector sizes of 0.5 mm square and 1.3 mm square,
respectively.
It is desirable to have the detectors cover as little area as possible, so
that the system will be less sensitive to angular misalignment.

From a purely electronic/mechanical viewpoint, yes, but from an operational
one, a larger spot is less sensitive to a speck of dust or a small defect in the
scale grating.

The enterprising builder might remove the quadrant detector from a CD
player. Personally I don't like fussing with surface mount packages.

Now on to getting patterns printed on plastic. Forgive me if this is too
elementary, but there may be some out there who don't know this trick. The
trick is to use the Linotronic (type) film printer at a printing service.

Total GARBAGE! You will be lucky to get registration of .01" over the
whole image! There will be all sorts of periodic and some non-periodic
errors. These things are made to set type for newspapers, not make
measuring scales with .0001" accuracy!

People try to make printed circuit boards this way, and are horrified
when the two sides don't match up better than .020" or worse over
a 12" wide board.

These print on mylar film very high contrast stable patterns. The machines
have resolutions from between 1200 and 3600 dpi.

Resolution is not accuracy!

They are used by Mac zealots, and only understand file formats that are
popular in the graphic arts world.

Plain PostScript.

Jon

From: "Elliot Burke" <elliot@...>

[long discussion on making read head snipped]
That was pretty good; I'm going to file that someplace where there's a
chance I can find it again.

I had the read head for my Teledyne/Gurley DRO replaced recently. They
used 2 _teeny_ photocells on the PC board. Not really a lot to it.

Finally, let me add my voice to those who think that a CD with RT Linux and
EMC on it made as simple as possible to install is a good idea. Like, put
in the disk, get a instant machine controller. Is this possible?

Not too likely with arbitrary hardware, unfortunately. It might work
if you said 'given this video card, any of these network cards, this
type of hardrive and CD', but most people wouldn't know the techy
details and it's hopeless taking a shopping list to a computer store.
Given the range of hardware out there, it's pretty amazing it works as
'painlessly' as it does (note the quotes around that). Heck, I have a
machine I'm working on now where Windoze won't even recognize the
mouse, serial _or_ PS/2. Probably a hardware glitch, but still a pain
...

How about a bundled system including the servo board as an option?
It's a lot easier to match up the software if you have the hardware
handy and it's the wide range of video cards available that makes
configuring X a pain. We end up building PC systems for folks locally
(Windows 9x with NT or Linux file servers, depending on what they
need). It's more interesting to build Linux systems.

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