Building Phantom Hardware and Firmware Mods

[jdcarpe]

litster, can you re-post the drawing file attachments for the Phantom plates? Or am I missing them somewhere? There was a .zip file with the drawings of all the plates. Also, I thought bpiphany (formerly PrinsValium) had posted the KiCAD files for the PCB...

[litster]

Where can I find this code?

on here: https://github.com/benblazak/ergodox-firmware (via http://ergodox.org/Firmware.aspx)

[litster]

One difference is that Phantom uses all Teensy pins, while ergoDox doesn't, as it has few keys, therefore fewer rows and columns are needed.

[Tranquilite]

So I've got basic function layer support working with one other layer that is active only when the function key is held down. This is only useful if you have a locking cherry switch installed, so I'm working on expanding this to something more useful.

[TheProfosist]

So I've got basic function layer support working with one other layer that is active only when the function key is held down. This is only useful if you have a locking cherry switch installed, so I'm working on expanding this to something more useful.

Thats fine for me as I have a HHKB esk layer. Also I think even Hazu's firmware that was used for the DOX was only able to do that also toggling layers without a locking switch wasnt possible

[Tranquilite]

Ok, so I've implemented toggle-able function layers now. I currently have 7 layers implemented (it should be fairly easy to add more) by pressing FN + 0-6, unfortunately, the way I programmed it lets you only map one function key.

I haven't really ever contributed anything source-wise before, so I'll just post my source here, and hopefully it shouldn't be too hard to get it incorporated into the main repository.  I'm pretty sure the only files I modified were phantom.c, phantom.h, and teensy_keyboard.c.

Some notes:
Everything that needs to be configured is located in phantom.c. The function key, as well as the keys needed to change layers are referenced by switch position (E.g. SW4_9 is switch where '0)' usually is). Mapping the layers is the same as previously, except that there is a layout array for each layer. Currently all of the layouts are the same ansiwin layout.

Umm. Enjoy! (Now somebody get media keys working )

EDIT: Also, if you want to have hold-function-key-to-switch-to-other-layer functionality, you can uncomment the other send() function in teensy_keyboard.c.

EDIT2: I've also found another problem with my function code. All layers use the same is_modifier array, so any key that is a modifier must remain a modifier (no switching caps_lock and control via layers). I'll see if I can find a graceful way of dealing with this.

[TheProfosist]

And my phantom is finally alive thanks to Tranquilite. It is definitely going to take some getting used to especially with the split up spacebars but i think it will come in time probably making me want a second one for myself so one can stay at home and one can come with me.

[CanadaRox]

Is there any special technique for installing Costar stabilizers?  The white hook part barely fits between the black arms when they aren't installed.  When I put them in the mounting plate they squeeze together too much and the key can't even move.  I've tried removing them and putting them back in a few times hoping it hadn't seated right but its always the same.

[alaricljs]

Use a micro straight blade screw driver and gently press the black plastic piece toward the plate edge so that it engages properly.  You may or may not hear a click as it seats itself.  The tolerance is very tight.

Oh yeah, by gently I mean as hard as you can with confidence that if you slip you won't destroy it or cut your hand open.

[litster]

The red areas need to be fully seated onto the plate.  Follow alaricljs's suggestion.


Also, the stabilizer clips are made of plastic, so the width could vary.  You can see the difference between these two in the picture below:


Besides making sure the clips are seated perfectly, you can also gently bend the clip with your figures to make the gap wider.  I did that to a few of mine and the wide keys don't stick any more.

There is one more trick up my sleeves if you still can't get it to work.  But that is a tiny bit extreme :-)

[TheProfosist]

Has anyone been able to program media and mouse keys? I really need them for my layout as it feels incomplete as it is now.

[WhiteFireDragon]

Are you guys interested in an aluminum metal case? If someone can provide the CAD file for something that fits or an exact duplicate of the current stock filco case, I might be able to get a friend of mine to make it. He works at a CNC machine shop.

I wish I knew how to draw it up on autoCAD.

[TheProfosist]

Are you guys interested in an aluminum metal case? If someone can provide the CAD file for something that fits or an exact duplicate of the current stock filco case, I might be able to get a friend of mine to make it. He works at a CNC machine shop.

I wish I knew how to draw it up on autoCAD.

Im good im just using a PLU-87 case for mine.

[TheProfosist]

New Phantom layout finalized and template for WASD made by hazeluff










Also just sent out to get a quote for the matching plate

[damorgue]

Are you guys interested in an aluminum metal case? If someone can provide the CAD file for something that fits or an exact duplicate of the current stock filco case, I might be able to get a friend of mine to make it. He works at a CNC machine shop.

I wish I knew how to draw it up on autoCAD.

Give me dimensions and I can finalize one.

[telepete]

For you cats thinking about doing a case, my Phantom is naked and my wallet doesn't have the 350 hit points necessary for that MKC one. I'm in.

[WhiteFireDragon]

Are you guys interested in an aluminum metal case? If someone can provide the CAD file for something that fits or an exact duplicate of the current stock filco case, I might be able to get a friend of mine to make it. He works at a CNC machine shop.

I wish I knew how to draw it up on autoCAD.

Give me dimensions and I can finalize one.

I'm gonna order a digital caliper to get accurate measurements. Right now I only have a standard ruler . Any recommended micro calipers? Are the $10 ones on ebay any good?

For you cats thinking about doing a case, my Phantom is naked and my wallet doesn't have the 350 hit points necessary for that MKC one. I'm in.

Not even sure the $350 MKC case fits the phantom boards and plates.

[damorgue]

I'm gonna order a digital caliper to get accurate measurements. Right now I only have a standard ruler . Any recommended micro calipers? Are the $10 ones on ebay any good?

The differences between the cheap digital calipers and the expensive ones are that the expensive ones are absolute and don't break. It may measure a bit more accurately but even cheap ones measure two decimal points in mm fairly accurately, well, accurately enough for us anyway.

With non-digital calipers, I'd say the difference between a cheap and an expensive one is less noticeable. They don't break. The $10 ones on ebay will probably do fine. I just checked, there are even $1.50 ones, lol.

[bpiphany]

Are you guys interested in an aluminum metal case? If someone can provide the CAD file for something that fits or an exact duplicate of the current stock filco case, I might be able to get a friend of mine to make it. He works at a CNC machine shop.

I wish I knew how to draw it up on autoCAD.

Give me dimensions and I can finalize one.

I'm gonna order a digital caliper to get accurate measurements. Right now I only have a standard ruler . Any recommended micro calipers? Are the $10 ones on ebay any good?

For you cats thinking about doing a case, my Phantom is naked and my wallet doesn't have the 350 hit points necessary for that MKC one. I'm in.

Not even sure the $350 MKC case fits the phantom boards and plates.

There has been CAD files for both the Phantom plates and PCB around. I don't know if that was before the crash, but I could remedy that =) Not exactly the same as Filco case measurements. So you may still want to take those anyway.

[TheProfosist]

For you cats thinking about doing a case, my Phantom is naked and my wallet doesn't have the 350 hit points necessary for that MKC one. I'm in.

if you need a cheap case just pick up a PLU-87 its a direct knockoff of a Filco.

[CanadaRox]

Use a micro straight blade screw driver and gently press the black plastic piece toward the plate edge so that it engages properly.  You may or may not hear a click as it seats itself.  The tolerance is very tight.

Oh yeah, by gently I mean as hard as you can with confidence that if you slip you won't destroy it or cut your hand open.

I managed to actually get the key installed, but I can't seem to avoid a fair amount of friction.  I can feel a little bit of resistance when lowering the key but it still gets stuck coming up if you don't release it very quickly.  With the switch removed the keycap moves freely with just the stabilizers but if I put the stem in the cap it causes a noticeable amount of friction on the stabilizers.

Are you guys interested in an aluminum metal case? If someone can provide the CAD file for something that fits or an exact duplicate of the current stock filco case, I might be able to get a friend of mine to make it. He works at a CNC machine shop.

I wish I knew how to draw it up on autoCAD.

I'd love to have an aluminum case but it really comes down to how much it is.

[alaricljs]

BTW - For anyone that wants an easy way to remap their Phantom, I built this.

[litster]

Firmware modders, please keep the firmware compatible with alaricljs's tool.  If you need to change the mapping, please let alaricljs know so matching changes can be coordinated. 

[Tranquilite]

Ok, so I forked the main phantom firmware, and have it available here. I'm new to using GIT, so I hope things are working just fine.

Since my last update, I've changed the firmware so that it no longer needs the is_modifier array (instead I changed the layout array to a 16bit integer array, and I set an extra bit on the modifiers so that I can do a bitwise comparison to determine if a key should be a modifier or not). I've also made it so that you can configure which function the LEDs are assigned in too, so you can now have your LEDs track num-lock, caps-lock, scroll-lock, compose or kana (though I have no idea what compose and kana are for).

As for comparability with alaricljs' tool, the only differences are that there are multiple layout arrays (named layout0 through layout6) which are now uint16_t instead of uint8_t, and there are no is_modifier arrays. Aside from that, I have kept the layout size and structure the same.

EDIT: I made some more changes, you can now use KEY_TEENSY_RESET in your layouts to jump to the teensy bootloader as if you had pressed the switch on the teensy. I'm sure this will be helpful for those who have their phantom in a case and don't want to open it up just to reprogram it. Also, I made it easier to switch between function key types (hold to change versus combination to change) and all you need to do is change the defined FUNCTION_TYPE in phantom.h.

[bpiphany]

I never really liked the is_modifier solution. It is a bit silly to have a huge array like that for the very few modifiers, although it is very time efficient... Merging it with layout is at least a bit tidier. I'm not too fond of the mess it introduces in other places, or the need for 16-bit operations. My experience with pointers is that they usually lead you to a dark, cold hole in the ground... But would it be possible to do something like this without getting yourself hurt too bad?

Code: [Select]

const uint16_t kbmap[NKEY] = {is_modifier[0]character[0], is_modifier[1]character[1], ...};
const uint8_t is_modifier[] = (uint8_t[])kbmap;
const uint8_t layout[] = (uint8_t[])kbmap + 1;
[...]
for(col = 0, key = 0; col < NCOL; col++) {
  for(row = 0; row < NROW; row++, key+=2) {
    [...]
  }
}

Or there may just be some assembler instructions to operate on the different parts of the 16-bit variables I suppose..

[Tranquilite]

Personally I find that pointers are extremely useful as long as you understand how they work and since arrays are just pointers to the first element of the array, I make it pretty easy to switch between layout arrays by creating a pointer that holds the address of the current layout array.

As for your suggestion for making layouts, I see that you are suggesting putting putting the contents of a layout array as well as the contents of an is_modifer array into a single kbmap array. I suppose it would work, but it seems a bit more complicated than the method that I implemented where each 16bit value in the layout is treated like two one byte values with the first byte specifies if it is a modifier or a regular key, and the second byte specifies what kind of modifier/key it is.

For example:
A = 0x00 04
Left Alt = 0x01 04

When my code looks at these values, it first blanks out the second byte via a bitwise and ( & 0xFF00) and then checks to see what type of key has been pressed:

A  becomes 0x00 00 == 0x00 00, so it knows that it is a regular key and stuffs 0x00 04 into the queue to be sent via usb_send()

Left alt becomes 0x01 00 == 0x01 00, so it knows that it is a modifier, so it blanks out the first byte with another bitwise and (0x01 04 & 0x00FF = 00 04)  and stuffs it into mod_keys.

While this seems kind of complicated, bitwise comparisons are very efficient as the compiler can usually deduce that only a few bits need to be flipped and checked. Additionally, this will be really useful for media keys and mouse keys (if somebody ever gets around to implementing them), as you can use the first byte to specify those as well.

The other benefit is that this is completely transparent to the user, as the layout[] array still looks exactly the same as it did previously (aside from being filled with 16bit integers instead of 8bit integers), and the only thing that has changed in the code are the definitions in usb_keyboard.h and the way that the program checks for modifiers.

[bpiphany]

I'm totally following your 16-bit values to define keys. That was smart =) Only the modifier definitions in the usb_keyboard.h file need to explicitly have the "is_modifier byte". The zero byte for all other keys is added automatically.

I may have been a bit shady in my explanation. I wanted to use your 16-bit key definitions in an array of 16-bit values. But with some pointer magic create two new 8-bit arrays. One that points to the most significant byte of the 16-bit array. Every other element in this array is a is_modifier boolean. The second 8-bit array points to the least significant bit of the first element in the 16-bit array, so that every other entry in this array is a character. (This of course is depends on if the MSB is first or last...) It is perhaps a bit of a dirty trick. But it allows for all 8-bit operations later on, and the check if a key is a modifier is a bit clearer. I don't think the average user needs to get bogged down in how that part of the firmware works. Setting up their own layers probably is plenty enough.

I agree, I love pointers screwing my mind up =D It is very useful to have full control, but that only happens once you actually know what you are doing...

[Tranquilite]

Oh ok, I understand what you want to do now, and I am pretty sure it would work too although you may have to typecast the 16bit integer pointer into an 8bit integer pointer in order to accomplish this. The thing I am confused about now is why we would want to do this. Doing all 8 bit operations later would be nice I guess, but you either need to make one 16 array work like two 8bit arrays (by incrementing by two each time), or go through the effort to make a bunch of new 8bit arrays. Though if we do this, we wouldn't want to treat the is_modifer array as a boolean because eventually I would like to have bytes to define media and mouse keys (though somebody else is going to have to figure out the USB HID enumeration for those, because I still don't understand how that works). I've also already defined 0xFFFF as the KEY_TEENSY_RESET to cause the teensy to jump to the bootloader without needing to press the button on the teensy.

I suppose my concern is that while this would make the modifier checks clearer, it would just make some other part of the code more complicated. I'll look into implementing it though and see which way I like more.

The problem with programming is not always finding a way to implement something, but rather finding the best way to implement it.

EDIT: One a side note, if anybody here has a better understanding of USB HID implementation, this document here seems to outline a pretty good way to implement media keys.

[bpiphany]

I tried to make up some c code in my earlier post with type casts and incrementing a memory address. I'm not quite familiar with those operations, I just know that adding a number to a pointer can be a bit confusing..

These lines just became a mess I think, and a bunch of extra 16-bit operations. But if you want to be able to use the most significant byte to indicate other things  like mouse or media keys, or whatever, I see how it can be hard to get around them.

Code: [Select]

if((layout[key] & 0x0100) == 0x0100)
  mod_keys |= (layout[key] & 0x00FF);


[Tranquilite]

Oh yeah, I forgot about that comparison. Right now it is only checking if the first bit of the first byte is set, and would obviously malfunction if I implemented something other than modifiers and regular keys. This can by fixed by changing the code to:

Code: [Select]

if((layout[key] & 0xFF00) == 0x0100)
  mod_keys |= (layout[key] & 0x00FF);
EDIT:Ok, so I made a couple functions that should work like the arrays that you were trying to create, but I haven't tested them yet so I am not sure if they will actually work.

Code: [Select]

uint8_t layout_value(uint8_t index){ //Just make sure that there are never more than 128 keys!
//create new layout pointer, type cast it to 8 bit pointer
uint8_t *layout_val = (uint8_t*) layout;
//index now needs to be twice as big plus one
index *= 2;
index++;
return layout_val[index];
}

uint8_t layout_type(uint8_t index){ // Just make sure that there are never more than 128 keys!
//create new layout pointer, type cast it to 8 bit pointer
uint8_t *layout_val = (uint8_t*) layout;
//index now needs to be twice as big
index *= 2;
return layout_val[index];
}
I suppose you could do this without creating a function for it by just making sure that your index is twice as big (and shifted by one if you want the 2nd byte).

[bpiphany]

Almost there =) What I want is to shift the layout_value pointer one byte. That way you can access both layout_type and layout_value with the same index variable. No incrementing every time you are accessing the value. The only funk is that you need to increment the index variable by 2 to advance one position. The 16-bit layout array is actually unused in the code. It is only there as a helper to facilitate tidier layout descriptions.

Code: [Select]

uint16_t layout[NKEY] = {layout_type[0]layout_value[0], layout_type[1]layout_value[1], ...};
uint8_t *layout_type = (uint8_t *)kbmap;
uint8_t *layout_value = (uint8_t *)kbmap + 1;
...
for(int row=0; row<NROW; row++, index+=2)
  if(layout_type[index] == interesting)
    do_something_with(index_value[index])


[TheProfosist]

Got my custom plates from the laser cutter today and their really nice and fit the switches perfectly.

[alaricljs]

What about the stabilizers?

[TheProfosist]

What about the stabilizers?

i guess i dont know what your asking about. I only have one key that needs stabilizers and that the spacebar I just picked some costar ones up from WASD.

[alaricljs]

fit the switches perfectly

^^ What about the stabilizer(s)  fitting?   Costar have really tight tolerances.

[TheProfosist]

fit the switches perfectly

^^ What about the stabilizer(s)  fitting?   Costar have really tight tolerances.

They fit in... I have yet to test with an actual spacebar though. I can once i get home if you want.

[TheProfosist]

fit the switches perfectly

^^ What about the stabilizer(s)  fitting?   Costar have really tight tolerances.

They fit in... I have yet to test with an actual spacebar though. I can once i get home if you want.

Their a touch tight and provide too much friction. With the spacebar on conventionally it wouldnt return up if you hit it in the middle if you hit it elsewhere there was about a 70% chance you would get it to return. Now it you put the spacebar on backwards which is the way i prefer it it actually works like 90% on the time. I think I may put a stoner spring in the spacebar and lube up the stabilizers and their clips and se it that makes a difference.

[litster]

are you using 6.25x spacebar?

[TheProfosist]

are you using 6.25x spacebar?

yep

[TheProfosist]

Also I swapped the spring out with a clear spring and at least backwards the spacebar is working 100% and is even snappy.

[damorgue]

Why are there no pics posted of people building these? I ordered diodes three weeks ago from ebay but haven't gotten them yet, so I soldered the teensy for now. I have sort of gathered parts on the side because I missed the original GB.

Here is proof that one can do it with a $5 soldering iron and a way too large tip. Don't mind the solder flux.
http://imgur.com/a/sfXSS

I can't wait for my diodes to arrive.

Edit: Thanks to a certain someone who generously gave up his Phantom to me.

[litster]

damorgue, you may want to make sure the teeny solder points are flush with the PCB.  Switches around they may not sit perfectly if the solder blob are above the PCB plane.  If you have switches on hand, try testing by putting switches in those switch positions around the teensy legs to make sure the switch sit flat on the PCB.

[damorgue]

Since I posted that, I received diodes as well as a flat side cutter(!). I used it to adjust them. I tried to check if the switch was flush before adjusting them and couldn't notice any leaning, but better safe than sorry.

Edit: I didn't know they were this cheap and of that high quality. I bought:
http://www.ebay.com/itm/ws/eBayISAPI.dll?ViewItem&item=190709158731&ssPageName=ADME:L:OU:GB:3160
It ships from Europe and not Asia which meant it arrived really fast and I am so amazed at how well it works. I began to search for one after I saw someone mention theirs in a thread: I forgot who it was though.

[TheProfosist]

Flat side cutters are o so helpful.

[damorgue]

My PCB is missing a hole, is that common for all the boards? I don't think those holes ares even necessary anymore to open the switches without desoldering. Lower left of the F11 key?

[litster]

Can you post a picture of the missing hole?

[damorgue]

http://i.imgur.com/NubN7.jpg
Lower left of F11. It does not have coating, but there is no hole. Are all PCBs like that?

Edit: I took the picture against a white surface in direct light to make it more noticeable. My desk is almost the same color as the PCB.

[litster]

I will have to check when I get home.  If you are talking about the yellow dot, yeah, you don't need it any more.

[TheProfosist]

That does look a bit odd...

[damorgue]

I would have guessed that they were all identical since the holes aren't drilled manually.