any thoughts on the new silouette portrait?

[thethereza]

No description provided.

[pmonta]

I think it's likely that the Portait will work in the same way, but since I don't have one, I can't be certain. Please try it out if you're able to. You may have to experiment with the offset setting along the x axis because of the Portrait's smaller span in that direction.

[thethereza]

I ordered one, was about $80 cheaper on amazon. They claim the internals
are the same so hopefully it will yield good results. I'll be using it on
windows. I'll let you know what how it works. Let me know if you have any
tips/suggestions.

-Reza

On Feb 7, 2013, at 4:05 AM, Peter Monta [email protected] wrote:

I think it's likely that the Portait will work in the same way, but since I
don't have one, I can't be certain. Please try it out if you're able to.
You may have to experiment with the offset setting along the x axis because
of the Portrait's smaller span in that direction.

—
Reply to this email directly or view it on
GitHubhttps://github.com//issues/1#issuecomment-13233040.

[thethereza]

Got it in, used vmware to run your code on a linux image. The drive was
/dev/usblp0 otherwise it worked as advertised. It was a bit difficult
to understand what the tests are supposed to do or what I should be
checking for. I've run a couple of them on some kapton that I have and
scanned the image at high resolution. You can download it here:
https://www.dropbox.com/s/19znxi1ec2b6fod/cuts.jpg -- let me know how it
looks compared to the output form your system.

It'll be a few more weeks before I need to make stencils with it, so
I'll wait till then before I play with it again. It's actually pretty
cool in what it does.

By the way, one annoyance is that the bits that get cut are stuck on the
adhesive pad. How do you remove them (given that there will be a ton of
tiny ones with a PCB stencil cutout).

Thanks,
Reza

Peter Monta wrote:

I think it's likely that the Portait will work in the same way, but
since I don't have one, I can't be certain. Please try it out if
you're able to. You may have to experiment with the offset setting
along the x axis because of the Portrait's smaller span in that direction.

—
Reply to this email directly or view it on GitHub
#1 (comment).

[pmonta]

Sounds good. I have a blog post with a bit more detail here:

http://pmonta.com/blog/2012/12/25/smt-stencil-cutting/

Your scanned image looks reasonable---it seems to be the output of a few of the test scripts. If you haven't done so already, you may want to try the gerber files in the examples directory (such as test_0.5mm_0402.gbr).

I've had good luck with mylar, and it's less expensive than kapton. The high-temperature capability of kapton may not be needed for stencil applications.

Cheers,
Peter

[thethereza]

I have a large roll of kapton that I was going to use in a laser cutter
for stencils but I never got clean cuts. I might as well use it now.

Where do you get your Mylar?

-Reza

On Feb 9, 2013, at 1:07 AM, Peter Monta [email protected] wrote:

Sounds good. I have a blog post with a bit more detail here:

http://pmonta.com/blog/2012/12/25/smt-stencil-cutting/

Your scanned image looks reasonable---it seems to be the output of a few of
the test scripts. If you haven't done so already, you may want to try the
gerber files in the examples directory (such as test_0.5mm_0402.gbr).

I've had good luck with mylar, and it's less expensive than kapton. The
high-temperature capability of kapton may not be needed for stencil
applications.

Cheers,
Peter

—
Reply to this email directly or view it on
GitHubhttps://github.com//issues/1#issuecomment-13328384..

[thethereza]

trying out the test pattern now, it's doing a lot of back and forth
without cutting much (it's still running). is that to be expected?
also, two questions - how do i compute your matrix from the calibration
test pattern, and why don't you shrink the pads inside the app vs.
trying to do it in the gerber - i'm not sure how to do it using the
gerber beforehand.

thanks
reza

Peter Monta wrote:

Sounds good. I have a blog post with a bit more detail here:

http://pmonta.com/blog/2012/12/25/smt-stencil-cutting/

Your scanned image looks reasonable---it seems to be the output of a
few of the test scripts. If you haven't done so already, you may want
to try the gerber files in the examples directory (such as
test_0.5mm_0402.gbr).

I've had good luck with mylar, and it's less expensive than kapton.
The high-temperature capability of kapton may not be needed for
stencil applications.

Cheers,
Peter

—
Reply to this email directly or view it on GitHub
#1 (comment)..

[pmonta]

Yes, the back-and-forth is to be expected---it's there to guarantee that a cut segment is always approached from the same direction, to eliminate any backlash in the positioning mechanism for better accuracy. It makes a noticeable difference, at least on my machine. You can try "--cut-mode 1" for faster speed, but you will see knife-drag distortion at the corners and backlash problems. Might be okay for large features (0805 pads and up). My tendency is to go for best accuracy even if I have to wait a long time (overnight say).

I should write a short note on computing the matrix, but briefly, "--matrix a,b,c,d" maps each point (x,y) in the gerber file to (ax+by, cx+dy) in device coordinates. So when you print the calibration pattern, if the distance between pattern groups along the x axis (the axis parallel to the main machine guide rods) is, say, 1% too large, then you should make the matrix "a" coefficient 0.99 to compensate. Similarly for the y-axis and the matrix "d" coefficient. That leaves "b" and "c", which control coupling between the two axes, which happens when the axes are not perfectly orthogonal. This is where the two calibration groupings at 45 degrees and -45 degrees come into play. If their distances differ, then you want to adjust "b" and "c" to compensate. To verify, print out the calibration pattern again using the candidate matrix; if you then get exactly equal distances in x, y, 45deg, and -45deg, and they're all equal to the intended distance (which I think is 5.00 inches on my example pattern), then you're done.

As for shrinking the pads, it did seem like the best place to do the processing was the PCB tool. It would be possible with just the gerber files, but I wanted to avoid parsing them; pstoedit removes the knowledge about pads and just dices them into individual segments, so it's then ambiguous in which direction to move the segments to implement the shrink. If your gerber file has no polygons (just apertures), which is usually the case for paste layers, you can edit the aperture header manually to shrink all the pads, if that seems preferable to using the CAM features of the PCB layout tool.

Cheers,
Peter

[pmonta]

I got my Mylar sheets, adhesive backing sheets, and adhesive residue remover from Amazon:

http://www.amazon.com/gp/product/B00006HOY8/ref=oh_details_o04_s01_i01?ie=UTF8&psc=1
http://www.amazon.com/gp/product/B00006HQEJ/ref=oh_details_o04_s01_i02?ie=UTF8&psc=1
http://www.amazon.com/Goo-Gone-Multi-Purpose-Cleaner/dp/B000VP7EDA/ref=sr_1_4?ie=UTF8&qid=1360988612&sr=8-4&keywords=goo+gone

Cheers,
Peter

[thethereza]

I could be mistaken but it seemed like the antibacklash mechanism made each
pass were it would approach from right to left then go right again equal to
the distance of the pad, then cut the line from right to left. It would
then continue left. Is that what it is supposed to do?

-Reza

On Feb 15, 2013, at 8:16 PM, Peter Monta [email protected] wrote:

Yes, the back-and-forth is to be expected---it's there to guarantee that a
cut segment is always approached from the same direction, to eliminate any
backlash in the positioning mechanism for better accuracy. It makes a
noticeable difference, at least on my machine. You can try "--cut-mode 1"
for faster speed, but you will see knife-drag distortion at the corners and
backlash problems. Might be okay for large features (0805 pads and up). My
tendency is to go for best accuracy even if I have to wait a long time
(overnight say).

I should write a short note on computing the matrix, but briefly, "--matrix
a,b,c,d" maps each point (x,y) in the gerber file to (ax+by, cx+dy) in
device coordinates. So when you print the calibration pattern, if the
distance between pattern groups along the x axis (the axis parallel to the
main machine guide rods) is, say, 1% too large, then you should make the
matrix "a" coefficient 0.99 to compensate. Similarly for the y-axis and the
matrix "d" coefficient. That leaves "b" and "c", which control coupling
between the two axes, which happens when the axes are not perfectly
orthogonal. This is where the two calibration groupings at 45 degrees and
-45 degrees come into play. If their distances differ, then you want to
adjust "b" and "c" to compensate. To verify, print out the calibration
pattern again using the candidate matrix; if you then get exactly equal
distances in x, y, 45deg, and -45deg, and they're all equal to the intended
distance (which I think is 5.00 inc hes on my example pattern), then you're
done.

As for shrinking the pads, it did seem like the best place to do the
processing was the PCB tool. It would be possible with just the gerber
files, but I wanted to avoid parsing them; pstoedit removes the knowledge
about pads and just dices them into individual segments, so it's then
ambiguous in which direction to move the segments to implement the shrink.
If your gerber file has no polygons (just apertures), which is usually the
case for paste layers, you can edit the aperture header manually to shrink
all the pads, if that seems preferable to using the CAM features of the PCB
layout tool.

Cheers,
Peter

—
Reply to this email directly or view it on
GitHubhttps://github.com//issues/1#issuecomment-13644179.

[thethereza]

The initial print looked really nice, though I just used the cutting
board attached and had to spend near an eternity trying to remove all
the little cutout pieces.

I was wondering how the scripts would handle circular/oval pads - given
your anti-backlash routines. Most of my pads are oval in shape:
http://i.imgur.com/kTwSdWK.png

Thanks,
Reza

Peter Monta wrote:

Yes, the back-and-forth is to be expected---it's there to guarantee
that a cut segment is always approached from the same direction, to
eliminate any backlash in the positioning mechanism for better
accuracy. It makes a noticeable difference, at least on my machine.
You can try "--cut-mode 1" for faster speed, but you will see
knife-drag distortion at the corners and backlash problems. Might be
okay for large features (0805 pads and up). My tendency is to go for
best accuracy even if I have to wait a long time (overnight say).

I should write a short note on computing the matrix, but briefly,
"--matrix a,b,c,d" maps each point (x,y) in the gerber file to (ax+by,
cx+dy) in device coordinates. So when you print the calibration
pattern, if the distance between pattern groups along the x axis (the
axis parallel to the main machine guide rods) is, say, 1% too large,
then you should make the matrix "a" coefficient 0.99 to compensate.
Similarly for the y-axis and the matrix "d" coefficient. That leaves
"b" and "c", which control coupling between the two axes, which
happens when the axes are not perfectly orthogonal. This is where the
two calibration groupings at 45 degrees and -45 degrees come into
play. If their distances differ, then you want to adjust "b" and "c"
to compensate. To verify, print out the calibration pattern again
using the candidate matrix; if you then get exactly equal distances in
x, y, 45deg, and -45deg, and they're all equal to the intended
distance (which I think is 5.00 inc hes on my example pattern), then
you're done.

As for shrinking the pads, it did seem like the best place to do the
processing was the PCB tool. It would be possible with just the gerber
files, but I wanted to avoid parsing them; pstoedit removes the
knowledge about pads and just dices them into individual segments, so
it's then ambiguous in which direction to move the segments to
implement the shrink. If your gerber file has no polygons (just
apertures), which is usually the case for paste layers, you can edit
the aperture header manually to shrink all the pads, if that seems
preferable to using the CAM features of the PCB layout tool.

Cheers,
Peter

—
Reply to this email directly or view it on GitHub
#1 (comment).

[pmonta]

I'll check whether the code and machine movements are doing the right thing for approach direction.

I haven't tried any oval pads, but my test pattern has some circular ones. I believe pstoedit makes either 8- or 16-sided polygons out of them. Any segments which are within the same 2*pi/16 bin I draw in the same "pass", since the knife has to rotate only a very small amount. Can you put your oval-pad gerber file somewhere so I can look at the output?

I should include a small verification tool to back-render the final graphtec output to PDF for ease in checking. Maybe include some small arrows to represent the stroke direction.

[thethereza]

my technique on how to deal with chads: http://rethinkmedical.com/cms/blog.php.

[aep]

Was there any conclusion on using the portrait? I got mine today and i'm very disappointed by the results. http://i.imgur.com/5Uch9dV.png Maybe the parameters everyone on the internet uses don't work well for the portrait.

[thethereza]

Mine works great. I use Kalyan film and you still have to manually Remove the chads. I make my own disposable backing using adhesive spray on a Mylar folder half.

On Jan 27, 2016, at 3:40 AM, Arvid E. Picciani [email protected] wrote:

Was there any conclusion on using the portrait? I got mine today and i'm very disappointed by the results. http://i.imgur.com/5Uch9dV.png Maybe the parameters everyone on the internet uses don't work well for the portrait.

—
Reply to this email directly or view it on GitHub.

[aep]

@thethereza can you share the params you're using for the portrait? (those are not chads btw, the stencil is completly ripped apart for fine pitch components)

[thethereza]

It's on a computer I don't use often. Pressure is 30, don't remember the rest.

On Jan 27, 2016, at 3:48 AM, Arvid E. Picciani [email protected] wrote:

@thethereza can you share the params you're using?

—
Reply to this email directly or view it on GitHub.

[aep]

@thethereza do you get more runs than specified with --force too? i.e. "--force 8" will run twice or three times

[thethereza]

i just do --speed 3 --force 30 (up to 40).

-r

Arvid E. Picciani wrote:

@thethereza https://github.com/thethereza do you get more runs than
specified with --force too? i.e. "--force 8" will run twice or three times

—
Reply to this email directly or view it on GitHub
#1 (comment).

[sergeyzwezdin]

@pmonta Hi Peter. I'm trying new Portrait device under Mac OS, but it doesn't appear in /dev/***. At the same time I can work with it at Silhouette Studio. Could you advise how to find mounting point in Mac?

[bikeNomad]

@sergeyzwezdin Mac OS doesn't mount the Portrait, because it doesn't present itself as a standard device class.
You need to use the file2graphtec spooler program, which will talk directly to the USB device on Mac OS.

I just cut a stencil on my Portrait (USB VID/PID 1123:0B4D) using it.
Here's what I did for a 5.7"x2" (overall, not component extent) PCB with the Gerber file origin set at the upper left corner of the board:

INFILE=6800_emulator-F.Paste.gbr 
./gerber2graphtec --offset 0,4 --media_size 8,12 --rotate 90 $INFILE > output.txt
./file2graphtec  output.txt

The resultant stencil (fed as media, not mat) ended up at (4,4) to (7.6,9.3).

I use Gorilla Tape to remove the chads from the sticky mat.

[sergeyzwezdin]

@bikeNomad Thank you! Very helpful!
@pmonta I believe it should be good to include this to readme file for Mac users