Fixing Signature Plastics SHIFT key legends digitally

[Oobly]

Sorry, some more off-topic chat:

The ownership issue can probably be covered by a legal letter giving explicit permission for "community" use, but it may get a bit difficult to word so it includes all GB's including those where small profit is made as in Ctrl|Alt buys, but not commercial use for SP's industrial clients, etc.

Another option is to give explicit decision-making power to a number of community members so if any of them give permission it can be used (and only one needs to give permission even if all the others say no). This should keep it usable even if the "owner" gets all huffy and doesn't want to share or if they go AWOL. Perhaps this group could be an amalgamation of the DT "club members" and GH moderators? That should be reasonably representative of the "community" and keep it fluid in the long term.

I'd prefer for it to not need to be a legal document as such, just a tacit agreement with SP to allow use for all community GB's, but they seem to be quite um... particular about covering their asses from a legal perspective.

It's good you brought up this issue as I'd like to get this kind of stuff sorted out before I start the fundraising proper so people know the situation very clearly. Well, back to Inkscape for some playing with fonts and stuff... sigh, not my favourite app...

[Muirium]

Beware: Inkscape's SVG files can show up badly in SP's (Old? Current? Who knows!) Adobe Illustrator. Constantly verifying what they see adds a lot of work. Try to get a copy of Illustrator instead.

[zslane]

Throwing another cat among the pigeons here, but the Vault 101 keycap from 00Zero's Out of the Vault Series 2 set has about 0.008" distance between the "1" and the "0"....

Do you have 00Zero's vector file to examine and confirm this? Because the naked eye is terrible at estimating such tiny distances. While those gaps may look small, I wouldn't be surprised if they are at least 0.020". The only way to know for sure is to open the file, isolate the actual outline paths and examine the point positions.

[7bit]

SHIF_T_vs_Yen_020164_724.jpg (76.82 KiB) Viewed 4649 times

SHIF_T_vs_Yen_020164.jpg (1.1 MiB) Viewed 4649 times

The ¥-symbol comes with thicker lines, because it is a larger legend.

As you can see, even within the SHIF T legend there are smaller gaps than between F and T.


Sorry for the dirt.


I've done the shot with a 105mm f/2.8 AiS with the PN-11 extension tube and flash SB-20 in manual mode, 1/125s at f/22. I did not bother to bounce the flash to make it look nicer.

[zslane]

Well, I've asked Melissa for a quote for our 25% legend applied to 25 keys, regardless of likely "voids" in the plastic. We'll see if my money is good enough to push this exercise through.

[obfuscated]

Why don't you just shorten both F and T with half the amount you've tried to shorten F?

About the ownership - make it public domain/non profit and don't bother with all the lawyer related stuff.

[zslane]

So I received a price quote from Melissa regarding a run of 25 keys with the proposed legend graphics. The amount was pretty much exactly what I expected. However, she asked that I amend the graphics to have those "center lines", which...I...just...don't...get. She also seemed to think I was giving her a TrueType font outline, rather than a hand-drawn recreation of their Gorton Modified letter glyphs. Clearly I am misunderstanding something very basic here.

So, I asked for a bit of clarification/explanation of this process and used the lotus flower legend as an example of my own expectations regarding custom legends with arbitrary shapes as outlines (no center lines). Our shapes may just happen to look like letters to a person, but to a machine they are just outlines with no special meaning. So hopefully she can explain this to me in a way that helps move this experiment forward.

[7bit]

I believe those graphics for Round 4 I've made where line graphics. The trouble was that the line widths came out extremely large in their buggy inlustrator crapware.


I've made them with texdraw. Not sure how to make line graphics with inkscape directly.

SP_alien.png (8.29 KiB) Viewed 4592 times

[zslane]

Well, the explanation I received from Melissa doesn't really help me. I guess I'd have to see the machines in action to appreciate the subtleties involved. Here is what she said:

This essentially boils down to the ability of the CNC machine and the size of the cutters available and their ability to 'do what we ask'. There are two ways a legend is cut, one by on-line - using a center line like I am requesting and one by in-fill - like the lotus flower in your example. The cutters that we use in our CNC machine really only go as small as 0.010 to 0.012 inches. When you are cutting an image that is really only 0.012 or 0.015 inches thick, the machine wants to overlap the tool path, which creates not only cosmetic issues, but often times breaks the cutter when going back over a pass it has already completed. The requirement for infill legends like the lotus flower is that the minimum line thickness be at least 0.020 inches, that way our smallest cutter isn't overlapping when the legend is cut.

Outline paths don't have a "line thickness" per se. In Illustrator I select "none" as the stroke and what I get is a shape with no stroke only a fill. The "outline" is nothing more than an imaginary boundary (with a thickness of 0) for the shape. All this talk of "line thickness" is terribly confusing. Even "shape thickness" is too simplistic for arbitrary shapes (like lotus petals) that are thin and pointy at the ends and fat and round in the middle.

Since I can't visualize what she's referring to, I basically have to take her word on this. I've sent her a version of the vector file with an additional layer containing the center lines, along with instructions to the effect that the stroke width of the resulting legend should match Gorton Modified (as used on the Dolch keys).

[7bit]

Not to me. She explained it very well. Why she did not do this much earlier?

Maybe there is an option to not fill the enclosed areas?

alienwild_fill_none.png (8.22 KiB) Viewed 4575 times

alienwild_fill_black.png (7.28 KiB) Viewed 4575 times

fill:none vs. fill:black

Code: Select all

<path
         d="m 140,1877.8 228,0"
         style="fill:none;stroke:#000000;stroke-width:112.80000305;stroke-linecap:round;stroke-linejoin:round;stroke-miterlimit:10;stroke-opacity:1;stroke-dasharray:none"
         id="path50" />

[zslane]

In an attempt demonstrate what I mean when I speak of shapes and paths, fill and stroke, I've traced the lotus flower (as best I could given the low screen-res version of it I had to grab from the Round 5a page).

Closed bezier paths describing lotus flower petals

lotus_paths.png (32.34 KiB) Viewed 4577 times

The flower is made up of eight separate pieces, each defined by its own closed bezier curve (called a "path" in illustration packages like Illustrator). They are not "lines". They are paths. Mathematical descriptions of boundaries. In this case, the eight bounded shapes are filled with black on the left and filled with nothing on the right. They have no outline stroke (it is set to 0 thickness). The red lines are merely Illustrator depicting the boundaries and the control points that create the shape of each piece (the "tangent handles" of each control point are not shown).

Notice how most of those petal shapes come to a very narrow point at one or more ends, and is fat and round in the middle. These shapes don't have a single "thickness" you can refer to, nor do they have outlines with thickness. They have an inner space (the positive space) and an outer space (the negative space) and an imaginary boundary line between them (shown in red).

[zslane]

Now, let's say SP uses a laser cutter in which the minimum kerf width that the laser spot is capable of achieving is 0.012", then I just don't see how you're going to get a lotus flower with petals shaped like the original Round 5a wiki drawing (with all its sharp end points and tiny gaps between petals). If, on the other hand, their "cutters" can product those lotus petals with such tiny details, despite Melissa's description, then cutting small letter glyphs like we have should be no problem, even with the in-fill method that will be used for the lotus legend.

[zslane]

That legend would be two shapes (one within the "island" of the other). There is no "line thickness" to deal with here either. Just two bounded areas (one of them being a small circle).

stickman_glyph.png (6.45 KiB) Viewed 4565 times

Of course, if the laser spot is so big that it can't trace those contours and spaces accurately enough, then you have a problem. But I get the impression that these legends with tiny details have been achievable in the past. I mean, just look at how the undersides of the arrowheads come to a sharp point (a "join"). That's simply not achievable without one hell of a small laser spot.

[7bit]

When I measure the trouch area of the row 3 SHIF T key, it is around 1200px and they measure 0.5" in reality, so 0.01" come down to about 24px.

The thin cutter breaks, becuase there is no material around it, except it touches the bottom of the legend mold.


@Melissa:
Why did you never give us such detailed specifications before!!!


Now, some things are even more clear to me.

I will post a picture of a lotus key quite soon, to show you that they used a thick cutter, which is thick enough to not break away when cutting and empty area ...


However:
The gap thing between SHIF and T is rubbish.
A smaller gap is possible!

[7bit]

That legend would be two shapes (one within the "island" of the other). There is no "line thickness" to deal with here either. Just two bounded areas (one of them being a small circle).

The attachment stickman_glyph.png is no longer available

Of course, if the laser spot is so big that it can't trace those contours and spaces accurately enough, then you have a problem. But I get the impression that these legends with tiny details have been achievable in the past.

Erhem!


There is no laser!

They engrave the legend into the tool!


Also:

alienwild_test2.png (5.34 KiB) Viewed 4559 times

Code: Select all

<path
         d="M 1275.2,850.602 1954.4,171.398"
         style="fill:none;stroke:#000000;stroke-width:7;stroke-linecap:round;stroke-linejoin:round;stroke-miterlimit:10;stroke-opacity:1;stroke-dasharray:none"
         id="path34" />

stroke-width:7 instead of 170 something.

Engrave with what? A metal drill bit?

Precisely!

[zslane]

Engrave with what? A metal drill bit?

[Ray]

About the ownership - make it public domain/non profit and don't bother with all the lawyer related stuff.

That wouldn't work, unless SP is fine with making no money…

On the tooling: sounds like drill bit to me as well, since laser beams usually don't break. But like 7bit I find it confusing it breaks when there is no material around it. Maybe it performs fatal oscillation on transition to already cut spaces

[7bit]

No, it is quite reasonable:

The drill bit is quite thin and touches the surface only with its head => no material around supports it => it breaks.

Will not happen with a thicker tool.

[Ray]

okay, kinda makes sense. Kinda sounds like bad design to me now, but I know nothing about tooling.

[Oobly]

SP use a CNC mill that uses end mills (which are somewhat like drill bits except the primary cutting surface includes not only the end, but the sides as well). These are made from a very brittle metal to be able to cut the metal of the mold and handle the high temperatures generated. The smaller the "bit" the more easily it snaps, especially with sideways force applied only to the end face.

...
I suspect there's been some miscommunication between the engineers and those who developed the specifications documents.

What I think is that the "centre line" thing is used for fast cutting of line font legends. Only the centre line path and thickness value are needed to do this and it results in a fast cutting process using an end mill with diameter the same as the line thickness.

"Block" legends (with a completely enclosed outside edge path) and images are likely cut with a much smaller "bit" and takes more setup, a much longer toolpath, results in more tool wear and takes a longer time to complete.

If I'm correct in this, most font legends can be cut using the simpler method (and should cost less), but some will still need to be processed as "blocks" (any varying thickness character).

...

And this does seem to be the case. Technically all they need for the SHIFT legend cutting is the centre line of the letters to use as a toolpath and the line thickness which will equal the diameter of the end mill they'll use to cut the legend.

The distance between parts of the letters themselves is only limited by the plastic flow characteristics and in this case I suspect you could have them practically touching without it becoming an issue since the ends of the lines are rounded, so I have to assume there's been a misunderstanding / misapplication of rules somewhere, since there is no practical reason for such a limitation.

What she is talking about is milling out a shape which has very thin lines inside the fill area, not outside. Outside lines can be as thin as the mold metal can stand during the molding process and whatever flow limitations the molded material has.

Shapes like the lotus flower will be cut using a very small end mill and spiral shaped toolpaths, starting in the centre of each "petal" area.

When she says "the machine wants to overlap the tool path", that is the result of a poorly generated toolpath and can actually be worked around by tweaking the G-code produced by the toolpath generation scripts / tools so the machine lifts the cutter during the pass. So even this is not a true limitation if you know your tools.

Sigh. I wish one the actual CNC engineers could be part of these discussions to eliminate all the confusion, misapplied rules and speculation.

To summarise:

1. Smallest radius on on any design = 0.010 to 0.012.
2. Smallest distance between parts of a design = material flow limitation. From what I've seen it can be very small (0.008) if the parts are close for only a small area / converge to the closest point and then diverge again.
3. Most characters with unvarying line thickness can be made using just a centre line and line thickness / tool diameter specified. This is the cheapest, quickest and most simple way to make legends.
4. Legends / designs that have varying line thickness will be defined by an outside 0 thickness enclosed path that will be used to generate the toolpath for a smaller diameter tool, based on the smallest radius in the design. The larger the tool can be the quicker and easier it can be cut.

I wonder if SP could just tell us what CNC tools they're using and let us generate the G-code. Would be simpler than all this back and forth with incompatible SVG files / PDF's and emails...

About the thicknesses on the Out of the Vault caps, I used large images and pixel counting as well as using a digital caliper on my own physical caps to measure the distances. The smallest distance is on the Nuka Cola cap, between the "C" and "o" at around 0.007" and it's actually on a support line, so it's a "difficult" spot on terms of flow, but the material flows in there just fine. Of course remelting during second shot molding can reduce the distances, but 0.020" is definitely larger than the smallest gaps on existing SA profile cap legends.

The rule would make sense if they were cutting "positive" versions of the legends, but for molds they need to cut "negative" versions, so it's just as I've said higher in the post.

So I propose generating a cente line only SVG of the SHIFT with more space around the I and less between the F and T so the overall distance after applying the correct line thickness is the 25% distance shown earlier and asking to pass it to the engineers for verification.

[7bit]

Thanks for your explaination!


I will carry this information into the wiki, as soon as I have time ...


Here is what I nailed together some years ago, with all the information I had back then:
http://deskthority.net/wiki/Graphics_for_insert_molding

[zslane]

I sent a version of the legend graphics with center lines on a new layer to Melissa (these are Illustrator files I've been sending her). I did not specify a stroke width apart from indicating I wanted it to match their current Gorton Modified SHIFT legend typeface thickness. The width is also, in essence, described by the outline path I originally gave her, which was derived from tracing a high-resolution scan of the original Gorton Modified legend.

I'm beginning to feel that Oobly is really the guy to be leading this effort...

[zslane]

One by-product of the in-line method is that because the cutting bit is round, all such line-based legends will have rounded ends, which is great if you want all your legend text to look like Helvetica Rounded. But what if you want your text to be a font with mitred ends (and usually at odd angles) rather than round ends?

I find it interesting that Oobly was able to get 0.007" gaps when Melissa insists that 0.020" is the smallest gap that won't produce "voids" when the plastic flows around the shapes. I hope he is right and she is wrong. 0.007" is the gap between the F and T in our 25% version (0.007" is 25% of the original gap which I measured to be ~0.027").

[7bit]

One reason I love double shots: rounded letters!


BTW:
May I have the SHIFT-SVG file?

If Melissa likes it, I like to compare it to my own files to see what goes wrong and make sure that my own creations receive the same love.

[zslane]

BTW:
May I have the SHIFT-SVG file?

If Melissa likes it, I like to compare it to my own files to see what goes wrong and make sure that my own creations receive the same love.

Of course, 7bit.

But keep in mind that I have been sending Melissa Illustrator (.ai) files, not SVG files. Would the Illustrator file still be useful to you?

Now, I can't really say if Melissa likes the center line version I sent her, but she accepted it and we're moving forward with manufacturing 25 keycaps as a test. This will be a $150 experiment (not counting shipping from WA to CA) for those interested in such details.

[7bit]

Only if it is easy to hack, or if I would be able to generate them with ImageMagick or GIMP.

Should I let one of my PayPal-slaves send you some money, or pay directly to SP?

Also: I hope you do 1 or 1.25units keys.

[zslane]

I will send you the .ai file and then if you can open it, you can export out any format you like I guess. I don't know why anyone would want to work with pixels though. Bezier curves are the right way to do this sort of work.

The point of this experiment is to see if they can even make what we want, and I'm using 2.75U Dolch DSA keycaps as the test case. If it works, then a real group buy can be organized by whomever likes doing such things for whichever keycap sizes and colors they want to make available. I would hope that standard Dolch DSA shift keys would be first on that list at least. I intend to not be involved directly in that part of it.

[Muirium]

DSA?

[7bit]

I will send you the .ai file and then if you can open it, you can export out any format you like I guess. I don't know why anyone would want to work with pixels though. Bezier curves are the right way to do this sort of work.

The point of this experiment is to see if they can even make what we want, and I'm using 2.75U Dolch DSA keycaps as the test case. If it works, then a real group buy can be organized by whomever likes doing such things for whichever keycap sizes and colors they want to make available. I would hope that standard Dolch DSA shift keys would be first on that list at least. I intend to not be involved directly in that part of it.

Did I say anything about pixels?


What I mean is: If the ai format is hackable, I might be able to generate it directly. If not, maybe there exists a tool to generate it from whatever source (ideally a PostScript file).

[zslane]

Did I say anything about pixels?

You mentioned ImageMagick and GIMP, which means you intend to work with raster images, i.e., pixels.