Cherry Jammer blog

[Wilba]

I posted in about what I was working on:

So rather than clutter that thread, I'll start a new one.

This is my stupid hacky experimental jammer keyboard using Cherry MX keyswitches.

Inspiration is from here: http://musicscienceguy.vox.com/

You can read up about jammer keyboards, the Thummer, Wicki-Hayden layout, etc. This is intended to be a jammer with Wicki-Hayden layout, but you could make it do any other layout, like Janko.

Do you have any more detailed pictures of your own modified Cherry MX switches or maybe some information about how you did it? I know the principle behind how it works but I'd love to see how you solved the problem. I might opt for velocity sensitivity and seeing as I'm going to have 192 keys to deal with... Well... :P

This is my first attempt. The more refined version uses a larger, U-shaped contact glued to the actuator, but is essentially the same.

The contacts on the side of the switch are two pins from a female header. These are easily pulled out from the bottom and look like a two-pronged fork. I glue these to the side using superglue, then cut one prong off each and bend the other one down. They are placed lower than shown in the above photo, so after the bend, there's clearance between the tip and the cylindrical part of the keycap which goes onto the keyswitch actuator.

The contact on the keyswitch actuator is made from the metal disc inside a typical 12mm tactile switch... it's slightly concave so when in the tactile switch, it's the part that acts as the switch spring and which makes the contact (shorts the switch pins). It's just the right thickness, thin enough not to interfere with the keycap, stiff enough to cut into a shape and not wear away. I cut the disc in half, then with the half-disc, first cut a notch to match the actuator, and then trim the other three sides. I gradually trim it to the exact size of the actuator - it's too small to do any kind of measuring and marking, so I do it by eye and then test fit it until it's right, or until I've cut too much away and have to start again. I then superglue it on the actuator, making sure you don't get any on the top, or up the actuator shaft. This is why I use a U shape now, instead of the thin strip in the first photo - the U shape means more surface area for the glue, and also makes it easier to align it when you're gluing it, i.e. the "perfect fit" means it won't slide out of position during gluing. What's important is getting it as big as possible, so the contacts have a place to touch it, yet it's not hanging over the edge of the actuator and rubbing against the hole (and getting pushed off during a depress).

In case it isn't obvious, you do all that with the switch disassembled, wait until glue has fully dried, then put it all back together, make sure the keycap won't rub the two top contacts, do lots of tweaking until you get good contacts (test using a multimeter). The spring of the Cherry MX will push the actuator (and the metal contact) against the other two contacts, and since there's some slack in the actuator shaft, it will "touch" against both the top contacts sort of evenly. I'm not really explaining it right... basically it works, if you tweak the top contacts so they are the same height and a little bit lower than where the actuator wants to be when not pressed.

It's not a very easy process, still at the experimental stage. There might be a better, easier way to do the same thing. This just happens to be the way I did it first, because it's what I had on hand, hundreds of tactile switches lying around, and a few female headers. It's not that cheap either, salvaging a bit of metal out of a 30 cent switch... but then maybe that's not too bad... Worst case guestimate, I've used 100 tactile switches for 96 keyswitches, so that's only $33. I'll pay that to add velocity to a keyboard.

Anyone wanting to do the same thing needs to experiment and learn their own method of doing it, how much glue to use, where to position the bits, esp. how to cut the discs, since cutting is done by hand. I mean, I could write up how I do it, but then if you follow that process, it might not work for you. It works for me because I'm crazy and have good eyesight and can cut 0.1mm slivers off a tiny piece of thin metal that I hold down with a finger. You might be half blind with fat fingers and my process would be useless. DIY! It's not rocket science.

In all likelihood, it would have been quicker and only marginally more expensive to buy an Axis-49, connect it to a Core32 and run my own firmware to get Wicki-Hayden layout, instead of requiring some PC software to remap the notes. That thing seems to have good travel, hex keys, velocity sensitivity, rubbery goodness. Why did I bother then?

What I'm doing is a hacked together experiment... I mean, it will work, it will have velocity sensitivity, it will even have blinken blinken.. the experiment is more to do with the ergonomics... does an isomorphic (jammer) keyboard need hex keys or round keys? can you use computer keyboard keys? does having the actuation point half-way down the travel (instead of at the bottom) make a difference to playability? does familiarity with PC keyboards translate into a shorter learning curve? does using a computer keyboard keyswitch, preferred by typists that prefer not to "bottom out" when typing, result in some advantages when used in a music keyboard?

I'll be happy when it's put together and works, I'm just not saying now that it's going to be anything special, or some great leap in keyboard design. It's just an experiment. At the very least, it will be something interesting to jam with and help me learn music without being held back by the classic piano keyboard layout.

[jaytee]

Did this ever go anywhere?

 

I currently use an Axis 49 together with some software magic to achieve a Wicki-Hayden layout. Now don't get me wrong, I like this set up a lot. I've never been much of a player-of-instruments, so it really just functions as a more-intuitive note input device and musical sketch pad, meaning my requirements are rather modest co pared to someone who is looking to play live keyboard parts.

 

That said, there's always been a few things about it that irk me and ever since I got this set up four years ago, I've been looking for alternatives, either DIY or commercial (and come up dry). The main things I want to address:

- The Axis is USB MIDI only. I would love to control hardware directly without a computer in between; normally I would look at a USB-MIDI host device, but because I rely on software to get a Wicki-Hayden layout, that won't work.

- The Axis 49, when turned 90° for Wicki-Hayden, is not quite wide enough to provide for optimal fingering in all keys. Right now it's set up in the best compromise layout, allowing good access to the most common scales, but there are a few scales with sharp or flat roots that are impossible to play without "wrapping" from one side of the keyboard to the other.

- No aftertouch or other expressivity beyond velocity. I feel like a grid layout would be even better for polyphonic aftertouch control than a piano keyboard; you don't need to stretch your fingers as far, so you may achieve better control.

 

@Wilba, as you mention, I could gut the Axis and install a MIDIbox core. This could at least solve the first problem, allowing me to define note assignments at the hardware level and add DIN MIDI. I even have a second Axis 49 (I bought two to achieve the ideal "jammer" layout, with the notes on the left-hand keyboard mirroring those on the right, but I find I rarely use the second keyboard in practice) so I wouldn't even be risking my prized, now-out-production main keyboard. However, that still doesn't solve the other two problems, which are honestly the most important to me.

 

I've looked into dozens of possible DIY solutions for making my own velocity-sensitive keyboard, and most of them are less-than-satisfactory. There's one person who built an isomorphic keyboard using a different Cherry modification technique (cutting off the bottom so the plunger can go through the PCB and hit a tactile switch on a second PCB underneath); this seems less than ideal. I've looked into having my own silicone membranes manufactured to do something like what is inside the Axis 49, but with the silicone as the actual button (like a TV remote or Monome); this gets expensive quickly, but could maybe work. And I looked into a company called Sensitronics who make sheets that when coupled with a properly designed PCB allow for any layout of FSRs you can imagine; I like this idea the best, but don't know what I would do for the actual mechanics of the keys.

 

I had the "brilliant" idea of using SPDT switches and timing the difference between the make and break...only to find this thread (which is how I ended up here) telling me that no such switch exists.

 

Le sigh... Anyone come up with any other ideas in the last few years? It's surprising to me that nobody manufactures a part that can do this.

[Antichambre]

Hi jaytee,

There's a lot of other idea on the web, I was looking for a velocity sensitive switch solution too.
The best I found is using hall effect sensor:
https://deskthority.net/workshop-f7/analog-numpad-with-hall-effect-sensors-t11191.html#p331469

But this solution is a little bit tricky because of the magnet.

 I'm already working on another solution which uses an infrared reflective sensor sharp GP2S700, always with cherry MX(red linear model ;). Today I just ordered a small pcb to test it. Can tell you back the result if you're interested.

BR
Antichambre

[jaytee]

Can you give a quick explanation on how your proposed solution works? I took a look at that data sheet but can't really figure out what it does.

 

either way, definitely interested to hear if it works!

[Antichambre]

Hi Jaytee,


Ok, it's simple, there's a infrared led and an infrared detector in the same package. The light will reflect on an object and come back to the detector. In fig.8 page 5 of the datasheet you will see the response between the object position and detector collector current. best response is for distance d in the range of 0 to 4 mm which is exactly the length travel of a cherry MX, now the 'object' where the light reflect is the moving part of the MX, you just have to cut the bottom centering plastic pin of the MX and place the detector under the center of the MX,
MX mechanical
The sharp detector works better with red object, you just have to buy red model which is a linear one.
See Mikroelectronika color demo
For electronic you just need a dual comparator chip to detect 2 separated threshold(one high one low) and timer counter between the two, it's the way I will follow in my test.

BR

Antichambre

Edited November 12, 2016 by Antichambre

[timefox]

fun fact,
C64 keyboard happens to be a matrix of variable resistors