wcass' Compact SSK: a review, reverse engineer, teardown, geek-out.

User avatar
Techno Trousers
100,000,000 actuations

04 Jan 2019, 23:33

I'd be fine with clamping a flat PCB between F-122 plates and heating it with a hot air gun or hair dryer until it takes on the curve. IIRC, that's what I$ did for the original FSSK.

The problem with strips are the interconnects between them. Lots of potential points of failure there.

User avatar
Muirium
µ

05 Jan 2019, 00:24

True. But from what I gather the FSSK had its own problems, quite plausibly from that hand bent PCB. I’d much rather have a kit of single row PCBs, even if I had to solder connections myself, than a single costly board I had just one shot to bend. The thought of waving around a heat gun to coerce the very heart of the thing gives me shivers!

__red__

05 Jan 2019, 04:02

Hypersphere wrote: The technology has existed for some time to make roll-up musical keyboards, and now foldable phones and displays are making debuts. Should it not be possible to use the same technologies to make flexible keyboard PCBs?
I just haven't found the lack of curve to be any issue. Try it before you overcomplicate it.

But, since we're talking about overcomplication:

You can do 0.4mm PCBs which may be flexible enough without needing to be abused - it's expensive.

You can do a real, genuine flexpcb, but now you're talking real money - especially if you go to 4 layers.


Lastly - given that capacitance varies based on the distance between the plates, the dielectric, and the thickness of the copper you're going to end up with significantly different analog values at the controller. You may want to poll DMA to check with him about potential firmware changes needed to adjust the range.

It's certainly not impossible, but it'll be spendy.

__red__

05 Jan 2019, 04:11

For 300x200mm (4 layer):
1 -> $300
10 -> $78
100 -> $25

User avatar
Muirium
µ

05 Jan 2019, 04:20

Easy: one CommonSense controller per row! And I name this sweet idea NonSense…

Anyway, I, for one, find spaced out caps (which were designed for a curved backplate, not a plane) to be an aesthetic and likely functional showstopper. The curve is there for a reason. Going flat transforms the character of the keyboard.

Depends how much you can stand flatness, I suppose. I like my caps cylindrical and row contoured, so, for me, IBM caps on flatness is striking to the point of a straight up no-go.

User avatar
Darkshado

05 Jan 2019, 04:59

Speaking of flexible PCBs, this popped up on my FB feed earlier today: https://blog.hackster.io/new-from-osh-p ... f343d6ebd3 I did not run the numbers for a typical keyboard though

__red__

05 Jan 2019, 05:38

Muirium wrote: Depends how much you can stand flatness, I suppose. I like my caps cylindrical and row contoured, so, for me, IBM caps on flatness is striking to the point of a straight up no-go.
To be clear, my daily driver is an F-122 at full incline so I'm pretty much typing at a 45 degree angle ;-)

User avatar
Sangdrax

05 Jan 2019, 11:35

depletedvespene wrote:
Sangdrax wrote: The only material upgrade I could imagine would be using a 4 axis mill to tool everything the same way but with the classic profile curve.
Would it be at all possible to make, for a lack of a better term, a set of PCB strips, all to be connected to the "main"
Yes. The only thing that matters is the pads themselves be aligned under the switches. You could even connect the strips via ribbon cables and then have the last strip connect to the controller. I've seen it done that way on Magnetic valve battleship boards which are just as finicky in sensing.

User avatar
Muirium
µ

05 Jan 2019, 18:29

__red__ wrote: To be clear, my daily driver is an F-122 at full incline so I'm pretty much typing at a 45 degree angle ;-)
When you're all hoist up like that, things are different, I'm sure. I use my boards flat on the desk (or my lap) so I'm very aware indeed of the flat ones vs. row profiles. As pretty as DSA can be, I always prefer sculpted caps on flat boards, and my IBMs as they were designed! Curve matters.

User avatar
Hypersphere

05 Jan 2019, 20:47

I'm in full agreement with Murium -- curve matters!

DMA

27 Jan 2019, 21:37

My soldering is so shitty LOL. But hey, it works.

SNR - wcass decided not to put vias between the floating pad and the pad behind it. With vias signal would be another 2x-4x stronger. Not that it's needed.. but I woudln't turn away from extra 6dB. If nothing else it would allow to drop ADC resolution and have even higher scanrate.
So, 0.4mm PCB is perfectly fine. Flex PCB is fine as well - but srsly, the cost of that is _obscene_.

User avatar
depletedvespene

27 Jan 2019, 22:18

DMA wrote:
27 Jan 2019, 21:37
So, 0.4mm PCB is perfectly fine. Flex PCB is fine as well - but srsly, the cost of that is _obscene_.
PCB strips it is, then? Let's remember that Model M keyboards (and, by extension, these Model W units) don't actually have a problem with keys taller than 1U (ISO Enter and the vertical 2U keys on the numpad simply use two barrels, one populated with the flipper and the other with a stabilizer).

DMA

28 Jan 2019, 01:29

depletedvespene wrote:
27 Jan 2019, 22:18
PCB strips it is, then?
God no. Strips will make much worse wiring mess.
0.8 is fine - model Fs still work after all those years and IBM didn't even care to bend them.
You don't need 4 layers unless you ABSOLUTELY must make it frameless (or plastic - so you must bring your own ground plane). Even then, if you're OK with wires connecting PCB to the controller - 2 layers is plenty.

As for the rollup pianos.. Technology exists to MASS-produce those. We tried to get a quote for mylar printed membranes not too long ago - they're insanely expensive even at hundred units.

Jampu

31 Jan 2019, 00:34

Have to agree with DMA - when I popped out an F122 sense card to hook it up to an experimental controller I was amazed to find that it's nothing more than a thin flat PCB retained by a couple tacks in the bottom plate. Most cheap fabs offer an option thin enough to flex easily, and given the form factor constraints (or, rather, the lack thereof) they should be ludicrously cheap. In fact, dropping the thickness to 0.6mm on ALLPCB dropped the unit price about 12%, bringing a 300x200mm board under $7 each shipped (at 25 order quantity). Drop in a $10 CS controller and there's your whole sense solution done.

User avatar
wcass

04 Feb 2019, 01:55

I went to the Workshop today to look up DMA's Bluetooth Common Sense and found this. Thanks red for the review.

I knew that a lot of folks would not care for the layout. I posted that back in September 2015 under the topic "Post a picture of your ideal keyboard layout!". I couldn't consider myself a real keyboard designer if i didn't make a keyboard with my own ideal layout now, could I. I plan on designing at least one keyboard a year because i enjoy doing it. To be honest, i have already started a new project (hence looking up info on Bluetooth Common Sense).

0.8 mm thick FR-4 is what IBM used and works fine with a slight bend (no heat gun required). As you go thinner, it gets easier to bend. 0.4 mm bends very easily. With or without a curved back; two layer or four layer; more design options is a good thing.

User avatar
depletedvespene

04 Feb 2019, 02:04

wcass wrote:
04 Feb 2019, 01:55
I plan on designing at least one keyboard a year because i enjoy doing it.
And you get 'em done, instead of just playing with KLE (like that git around here lacking a certain gas). Kudos!


(damn, I need to get way, way better when it comes to ass kissing) :mrgreen:

Post Reply

Return to “Workshop”