Fixing Signature Plastics SHIFT key legends digitally

[zslane]

So in summary:

It sounds like Muirium and 7bit are advocating that I just send the 25% or 50% version and see what they can do with it.

Others such as chzel appear to be advocating that I merge the F and T into a single glyph shape (but with the same distance between vertical stems as the 25% or 50% version).

Do I have that right?

[7bit]

Yes.

I know what SP fears: If it is a difficult legend, they will produce more trash, ie keys where the legend did not come out as intended. But I know that is is possible to have smaller gaps than just 0.02".


Also, I will cover the costs for the experiment.

[zslane]

In light of 7bit's confidence, I will send both the 25% and 50% versions and see which one comes out the best. We always have the merged glyph approach if both of those should fail completely.

[chzel]

If their process is as I imagine it, they use mold pieces akin to the movable types of old and assemble them to form a word.
So each letter has a minimum distance from the next based on their actual dimensions. Specifying a smaller distance would be of no use since they can't jam them closer to each other, unless they create a new F or T with a narrower base.
So, instead of letting them improvise in this step, we specify a shape that is the ligature "FT" with whatever separation is best, and have them treat it as any other shape (like the lotus or the kbd runner) just scaled to match their font size.

[zslane]

I don't know the specifics of the process, but they can clearly mold arbitrary shapes, like the duck or the kbdrunner or the lotus blossom. Look at the lotus blossom carefully. There are eight distinct shapes in it. Our shift key legend has only five, and the space between them won't be much different than what you see between the petals of the lotus blossom.

[chzel]

That's exactly my point, and that's why I prefer the "arbitrary shape that happens to look like a word" approach.
If we need to do the whole SHIFT as on shape or we can just do the FT, I don't know.
I'm probably not putting my thoughts to words clearly enough, but English is my second language, and I'm a bit tired (it's past midnight here)! Sorry for any confusion I might have created!

[Muirium]

We're talking about pretty esoteric stuff, so confusion is well excused!

Here's how I understand it. SP has two independent ways of creating doubleshot legends:

1. They have a full alphabet of ready made symbols in *non* *digital* *form*! They bodged it together in ancient times, and call it "Gorton Modified". They have no access to the people or techniques that created it. Gorton is simply a thing, as it is, never to be changed. Or at least that's what they say. We can make arbitrary legends (like DUCK and 7BOT MODE) in this alphabet, but have no control whatsoever over the kerning, which is of course complete SHIF T. As far as I know, the last new character was the € sign, so they are possible*. That would be the ideal way to fix this, with a igature.

Think of Gorton like this, because I don't think it's so far off:



Note: ligature! And it's two-piece! Just like we want.


2. They can create new custom legends, using vector files in Adobe Illustrator apparently. These the ones we can dicker about with. But there are different rules here. Firstly, the nonsense about 0.02 minimum tolerances. This is a flawed understanding on SP's part. They've broken that rule just fine before, but they're real good at forgetting. Secondly, custom / vector legends are subject to wild variations in weight (boldness) and can get messed up badly when moved between rows.

Vector legends can, and do, turn out like this:



Same. Freaking. Vector. I need a drink…

Facet, who worked with SP on doubleshot ducks and those hashbangs, knows of what he speaks:

About the SHIFT, I'd be wary of going down the SVG route.

As has been mentioned, SP use a version of Adobe Illustrator (and I don't) and for Honey Sphericals Extras I was finding scaling problems with exports from Inkscape which I could never get to the bottom of (and I tried enough iterations to annoy both me and SP - I never verified a single factor...) We got around this by providing a supporting PDF, quoting the dimensions, and had a screenshot sent back showing the rescaled and correct dimensions in Illustrator. We had to do this for the duck because to SP it seemed to be going over the legendable area boundary. We got a great result in the end (thanks matt3o!).

The #! (hashbang) legend didn't get any such check though, so it was waved through, and production messed it up... twice... The legend itself was intended to match the tall Round 5 alphas, but it came out a bit thicker than intended on the row 3 key. When they remade it as the row 1 key it was supposed to be, they upped the thickness dramatically because the tooling was recut on a different machine...

The other thing about submitting an SVG is that the tooling made is likely to be bound a single key size and shape. This means multiple legends would need to be paid for and means more scope for errors.

Note the bit about tooling being recut on a different machine. All this stuff is a hands-on, manual process. We should really have someone physically there, who can verify things in place, instead of the expense of long distance iteration. You don't fancy a drive up to the Canadian border do you, Zslane? SP's up in Custer, Washington…


*Of course, the Euro sign might just be a vector legend for all I know. It's not exactly the kind of letter to show up in the middle of PAGE DOWN. I don't know what the rules are for the double decked legends like the number row keys. But I do reckon the two types cannot be mixed on a single key.

[zslane]

I know where SP is. They are a 35 minute drive north of my mother's house. Alas, I won't be up that way until Christmas.

[0100010]

Not sure if this will apply or help at all - but this Shift legend thread brings to mind an issue "we" had with the Nuclear Data Green SA Row 3 GB's right Shift key. The legend for it was extremely squished compared to the left Shift key - which you can see in this image :



The fixed right Shift key legend can be seen in this post : https://geekhack.org/index.php?topic=61 ... msg1535658

I only mention this in case it helps with this letter spacing discussion you are having with SP.

[zslane]

That looks like a mismatch of typeface size to me. Which only serves to demonstrate that the machine that creates molds from the metal type blocks can work just fine with very narrow lines and gaps if that is what a particular metal block series calls for. The machine that makes molds from digital vector files may have a completely different set of tolerances and limitations, and the two may not be at all comparable. We shall see.

[bloo]

50 is really very nice. 25 is gorgeous. I'm interested to hear how receptive SP is to this.

[7bit]

The gaps in the nuclear symbol are smaller than the distance between F and T.

[zslane]

It is very hard to tell from a low-res photograph like that though. Looks can be very deceiving. To my eyes, the gaps in the nuclear symbol near the shift key appear smaller than the gaps in the other nuclear symbol. Is that a sign of the inconsistency to come? I'm sure Muirium would emphatically say yes. But it does give some measure of hope, I suppose.

[Eszett]

To kern F and T is basics. And this shows that SP seems to not care about Kerning, not even about the basics. Look how unprofessional this company is. It is a tragedy. The only consequence can be to not buy from them anymore! At least this is my decision.

[Muirium]

Very smart decision. After Round 5, I'm more or less there with you. Can't really see another set from them that would appeal to me now. (Not even Alps.)

I'm sure Muirium would emphatically say yes. But it does give some measure of hope, I suppose.

I can't see the future. But they could roll a natural 20 I suppose. Or take 20+ tries to get there…

[7bit]

To kern F and T is basics. And this shows that SP seems to not care about Kerning, not even about the basics. Look how unprofessional this company is. It is a tragedy. The only consequence can be to not buy from them anymore! At least this is my decision.

It is your task to make sure QWER keys produce all the legends we ever need, with acceptable MOQs and at acceptable prices.

There is no other manufacturer that comes to my mind that would be able to produce sphericals.

SP is quite OK. They have their limitations, but they can do exceptional things, like all of those single legend keys, which (except for the Yen key) came out perfectly, without any need to submit them SVG graphics.


Make an SVG with the F and T very close to each other and tell them to scale it such that it fits with the medium sized legends. They have 3 sizes big ass, medium and condensed.

I have forbidden them to use big ass anymore, because they are too large, and look disharmoic next to smaller legends.

[Muirium]

QWER shut up shop, as I recall. You'd have to talk with Devlin. But only after some MOQsplainin' with GMK.

Can't we all just get along and dyesub?

[zslane]

Tonight I sent the Illustrator file to Melissa with the 25% version. The process has begun. Let's all pray it gets past her (likely) reluctance to try something with such minute gaps.

[XMIT]

I know where SP is. They are a 35 minute drive north of my mother's house. Alas, I won't be up that way until Christmas.

Get a factory tour and post lots of photos!

[zslane]

BTW, I just noticed that the space between the F and T in SP's standard dyesub PBT DSA shift keys very closely resembles our 25% solution.

[Muirium]

Hey Zslane, I finally scanned my caps. A bit on the late side, I know. But I scanned SP's SA and DSA doubleshots side by side. Here's a very small (but still quite hefty) version of the PNG file:

Tiny SHIF T.png (2.12 MiB) Viewed 4597 times

They just don't kern those Ts at all!

If you want the 151 megabyte original (already run through pngcrush etc.) I can share it via Dropbox.

Note how they are similar but not quite the same. The RETURN legends are especially distinct, SA vs. DSA. These caps are DSA Dolch and Round 4 SPH.

[zslane]

If you want the 151 megabyte original (already run through pngcrush etc.) I can share it via Dropbox.

Yeah, that would be ideal. And if possible, could you double-check that the original scan dpi is set for the image? Thanks!

[Muirium]

4800 DPI. Uploading to Dropbox now. I tweaked the levels a bit to increase contrast / look more realistic. No compression lossy though or resizing / rotating. There's a lot of plastic detail (little bubbles?) when viewed at full resolution. And blur, thanks to these being sphericals, with the centre of the legends a millimetre or two away from the image plane.

[zslane]

Wow, okay. I'd be satisfied with 1200dpi if you want to make the file size a bit more manageable.

[Muirium]

Ah, it'll be uploaded in half an hour or so. The scanner's away now.

[7bit]

Ah, it'll be uploaded in half an hour or so. The scanner's away now.

JPEG:

Tiny_SHIF_T.jpg (266.67 KiB) Viewed 4546 times

If there are many colors, JPEG is the way to go!!!


The file is only 10% in size of your PNG.

[Muirium]

Zslane wanted lossless. This is the real file:

https://dl.dropboxusercontent.com/u/361249/SHIF%20T.png

Get it while it's hot. I'll delete it from my Dropbox tomorrow. Not much space spare!

[zslane]

Got it. Thanks!

[Muirium]

Cool. I'll leave it up a while for 7bit to get his copy as well. I'm sure he can compress it into a sub megabyte jpeg too!

[zslane]

A comparison via image overlay shows that all of the letters on the blue SA shift key, with the exception of the S, are identical in shape and spacing as my Dolch DSA shift key. The SA key's S is slightly compressed vertically compared to my Dolch DSA key.

White legend is DSA, red legend is SA

DSA-SA-shift-key-legend-comparison.png (336.61 KiB) Viewed 4525 times

So the question is, what accounts for this difference in just one character? Is this due to the difference in top surface curvature? Is this an example of the variability inherent in SP's manufacturing tolerances? Maybe a combination of both?

Now I can re-trace the S from Muirium's scan and submit a new vector file for SA keys with that version of the S, but we face the prospect of it becoming further compressed if what we're seeing is a curvature artifact. Or I could submit the same vector file as the DSA key and go with the idea that it would look like the scan once it is used on the SA key.

Thoughts?