Can we design the teensy alternative for keyboards?

[matt3o]

the mid-mounted usb sticks below the PCB so it is not an option since you couldn't put the board flush over another PCB. So, really, not worth wasting more time on this.

the best we could do is to offer a USB-less version of the board unless there is some magical component I'm not aware of that solves all our problems

[vvp]

The connector mohitgarg proposed will not stick below the bottom PCB side if the PCB thickness is at least 1.2 mm.

Edit: Btw, that connector would stick 1.815 mm above the top side of the PCB. And it requires at least 1.125 mm thick PCB so that it does not stick below the PCB bottom surface.

[matt3o]

I was looking at Matt_'s schema



I thought they were more or less the same. If we can go down inside the PCB and still keep the island, I'm fine with that.

[Matt_]

Keep in mind that proportions are way off in this drawing. If the micro USB connector was 3 mm high overall (which is the standard height), the PCB shown here would be 0.6mm thick.

We'd have to check the datasheet of the particular USB connector we would use, but with a standard PCB (1.6 mm thick) it's very unlikely that the connector will stick out from the bottom. vvp's numbers sound right to me.

If we go this route, the Molex 47642 may be a good choice: http://eu.mouser.com/ProductDetail/Mole ... 6skg%3D%3D

[mohitgarg]

@Matt_, Yeah, the Molex 47642 looks good if we decide to go that route. I'll play around in KiCAD this weekend to see where we can go with this.

@flabbergast, That is very interesting to know how the firmware uploading will work on the ARM chip. I think it's a good time to start working on the bootloader, I'm hopeless when it comes to those kind of things. The architecture is set in stone, we're just working on the physical aspects now.

@matt3o, now that we have some space, do you think we should implement some USB protection circuitry, like a fuse and some TVS diodes for the USB data lines? The Teensy for reference does not use anything, however I have seen some cases on GeekHack where people seem to have fried their Teensy due to ESD (Mostly heard about the Ergodox).

[matt3o]

I'm totally okay at adding protections and more circuitry, just... the more components = more expensive board. So just try to find a good midway.

[pomk]

Hi all, I've been following this thread closely and I wonder if something like this could be useful?

demo.png

This is an interesting concept. What are the exact dimension of this board? You would also need to consider other switch orientations (namely upside down ) and compatibility with alps.

The board is exactly 1U x 2U, so 19.05 x 38.10.
With the through hole variant of the mid mount connector (thanks Matt_ for pointing out) I think that the routing will be a bit easier as the footprint is considerably smaller, it should also allow for upside down switches and alps switches without clearance issues. The mid mount connector sits flush with 1.6mm pcb bottom. The pcb connector can be moved up a bit if you want an island, although the cable connector already has a clearance of 2mm if I recall correctly and the pcb connector stick out by 1mm from the pcb outline. I can try to make the footprint for the through hole variant and see how much room we would have for routing.
I think that talking about islands is not that helpful, but rather what the compatible bezel sizes should be for a layered case.

[matt3o]

to sum-up. if we can keep the current shape footprint with a mid-mount USB port I'm totally fine with that, otherwise I would probably stick with a standard connector.

Also, I have a feeling the USB-C connector would solve all our issues... what do you think? Were we too hasty at dismissing it?



Pads in the middle for the switches is a nice idea (especially for hand wired keyboards). Is pomk's design worth exploring? If you remove 4 pads from the bottom you can probably have a small USB island. The removed pads could be placed back to the bottom on the longest side probably.

[mohitgarg]

The problem with mid-mount and an island is that the island will actually be two islands with dimensions of about 2.5*2.5. Not very strong.

USB-Type C is wider and taller, so in that aspect it provides no advantage. Not to mention about $1-1.5 more expensive.

I think it will be impossible to have mounting position for two switches and two mounting holes, this is because the mounting holes will have to be in the middle of the two switches and that is the only viable place for the MCU chip.

[Muirium]

USB C is a significant draw for this project, all by itself. I've scuffed up so many other USB ports because they're non reversible. C is the first USB that doesn't suck!

Hell, I'd like a USB C mod for my HHKB. It's that compelling.

[mohitgarg]

USB C is a significant draw for this project, all by itself. I've scuffed up so many other USB ports because they're non reversible. C is the first USB that doesn't suck!

Hell, I'd like a USB C mod for my HHKB. It's that compelling.

Tried the MicFlip cable?

[Muirium]

Nope. The concept alone is enough to make me angry! Matias "reversible" USB cable on the Ergo Pro got me that way too. "What the fuck is WRONG with this cable?" It didn't even fit in my MacBook Pro. Argh, don't get me started!

[matt3o]

I think it will be impossible to have mounting position for two switches and two mounting holes, this is because the mounting holes will have to be in the middle of the two switches and that is the only viable place for the MCU chip.

sorry again for being so dumb, but the mounting holes are not needed if you use the board directly on the switches, so the central switch hole would be used for either screws or to mount the switch.

[mohitgarg]

I was the one being dumb. That's make so much sense.

[Matt_]

I have mixed feelings about USB-C. On the one hand, it's the latest standard, cables and connectors are now available for relatively cheap, and it's reversible.

On the other hand, it makes the controller more expensive, takes more real estate on the board (wider & taller as Moz points out), and we don't even know if a significant proportion of the controller's users will actually use it directly (as opposed to using a separate connector for more flexibility).

As much as I love the idea of implementing type-C, I'm still leaning towards staying with Micro-USB. Which is less cutting-edge, but probably more practical.

[matt3o]

if USB-c doesn't help us in any way, I wouldn't use it just for the sake of it

[Muirium]

Aye, I hear you. At least Micro is better than vile, shitty Mini.

Speaking of which: USB C mod for HHKB is a serious request. I'd love that! And I wouldn't be alone. Especially if it was the magic Bluetooth unicorn that PFU/Topre just so failed to launch themselves.

[mohitgarg]

Why not take Hasu's project and swap the USB on it? It already has Bluetooth as well.

[vvp]

Pomk's design looks appealing but uses pin headers with 2 mm pitch. They (and the corresponding receptacles) are a bit harder to get than 2.54 mm pin headers. Pomk's design also needs a through hole USB connector.
Isn't Pomk's board wider? Will it fit under 6.25u symmetric space bar? Not that I care about fitting under space bar at all, but it was the initial requirement.

[mohitgarg]

I think matt3o is referring to in particular is having the mount holes for two switches on the current iteration

[Muirium]

All respect to Hasu (TMK is great and he's a sterling chap) his BT HHKB controller has impractically poor battery life. He bitbangs the USB protocol, as I understand it. Which uses an order of magnitude more power than native BT capsense, like Topre's own.

[vvp]

If it is only the switch hole being the same as mounting holes then that is a good idea (if you can fit controller in the middle).

As for as PCB tabs being too weak, well you do not need to make them as narrow as possible or you can use right trapezoid shape. If the USB peninsula advocates are OK with it. I think they need to decide whether they want to fit the controller between PCB and the mounting plate (which means a mid-mount connector) or whether they want a narrow USB peninsula.

[flabbergast]

My opinions:
- USB C: might be interesting to try to fit. The only disadvantages are size and more trouble for DYI-ers. But I'd be OK with that if we can fit it in. Don't care about price ($1 - $2 extra is fine, and it will be an excellent selling point).
- 2mm headers: this pretty much means we don't ever expect anyone to actually use it with pin headers/jumper wires/breadboard/with-non-custom-made-pluggable-boards. Which is more-less OK by me for this board (as it's really not the intended usage), but you should realise that it's very expensive, annoying and difficult to get 2mm-spaced things for end users. We will need to (well, really should) provide pin headers with the board, so that it can be at least mounted to a PCB without a 2mm-header-hunt.
- peninsula: I'm still for, if mechanically/size-wise viable. Still thinking custom cut sandwich cases (the cheapest case to have custom-made).
- pomk's layout: basically OK, if it will fit / can be routed. My personal minimal requirements are
1) careful placement of USB lines' pads (best would be between the USB connector and MCU, or at least not very far).
2) routing: basically OSH Park's design rules. The reason is that they're by far the best source of (usually) quality PCBs in very small quantities (for DIY people). Means 6mil trace width, 6mil spacing (the larger we can do the better; power/gnd lines at least 10mil), 13mil via drill, 27mil via dia. {The vias are a bit bigger than what the chinese fabs usually require, but the big advantage is that even if the drill hits are slightly off it's still OK.}

[pomk]

I was not able to route it using the OHS design rules without creating a small island

Should have more than enough space to go with 2,54mm headers as well.

demo2.png (292.2 KiB) Viewed 5975 times

Note that this was just a feasibility study and the board should be routed with a lot more care than what I gave it now.

[flabbergast]

BTW the three 0.1uF caps are supposed to be physically close to the three VDD/VSS pairs of MCU pins (that's why there are three caps).
Now when it looks like there will be space: for the 2.2uF caps, if 0603 footprints will fit, it would be better. These are caps that stabilise the voltage, one for the regulator output, one for the usb input. These caps should be the best (parameter-wise) we can get, and a bigger footprint means we can easily get caps with better voltage rating. {Voltage rating: the max voltage a cap can see and *not* blow. The bigger the better for these caps, to survive potential voltage spikes.}

EDIT: More comments about the onboard caps: the board *won't work for sure* without at least the 2.2uF on the voltage regulator output. It will probably work without the 0.1uF ones, although reliability may vary. The 2.2uF between USB 5V and GND is there to protect the MCU from spikes on USB (can happen e.g. when plugging in).

[mohitgarg]

Wait, I count 34 pins, don't we need like 41 pins? Also, I think we should have in-switch LEDs as well as diodes for the switches, or am I missing something here?

[Matt_]

I have to admit I don't quite understand the purpose of the switch pads. For builds with no PCB, where the controller is meant to be soldered under switches? Are we not losing lots of space (and possibly pads along the edges of the board) going that way?

[vvp]

Cumulative capacitive loading on USB power lines should not exceed 10 µF. That is the limit of the USB specs. If I understand it correctly we are now at 2*2.2 + 3*0.1 = 4.7 µF. So there is still space to increase the capacitors but not more than about 2 times.

[matt3o]

Wait, I count 34 pins, don't we need like 41 pins? Also, I think we should have in-switch LEDs as well as diodes for the switches, or am I missing something here?

technically we would also miss ALPs compatibility

[pomk]

Wait, I count 34 pins, don't we need like 41 pins? Also, I think we should have in-switch LEDs as well as diodes for the switches, or am I missing something here?

technically we would also miss ALPs compatibility

Just a feasibility study. Also no need for diodes, as those would be in the main pcb, for hand wired use I would add small holes near the pads so that you can easily wire them up to the whole matrix. Upside down alps is easy to add, for both orientations one would need to move the usb connector up a bit.