New Plate Idea?

[0100010]

Most keyboard plates and PCBs, are flat.  But to create a more natural typing slope, the key height and angle are profiled differently for each row.

Example :



But - in the case of some keyboards, the keys are all the same shape / height / angle, and it is the plate that is curved - like most buckling spring boards :




Was wondering if one could layout a keyboard plate for Cherry switches, such as :




But make it essentially curved.  Where each row would be a flat surface, but in between the rows would be a 3 or 4 degree bend.  That way, you could use keycaps that were all the same profile, like this :

Top = 4 degree
Middle = 3 degree
Bottom = 0 degree



Obviously the keys that span more than one row, such as the numpad + and Enter, would have to be cut as a separate plate, bent by itself and attached to the rest of the plate after.

Also - it doesn't lend itself well to having a PCB, so direct wiring would be easier.  Creating a case would be some work as well.

Thoughts?

[jdcarpe]

I personally like the idea. I would just hope that the switch hole cutouts don't get distorted when the plate is bent for the curve.

[0100010]

Yeah - that would come down to plate material / thickness and maybe the brake used to create the bend.  Would also need to increase the vertical spacing of the rows compared to normal, to allow for both bend radius and keycap clearance.  But I have no idea how much additional spacing would be needed.

In short - someone that knows sheet metal better than myself, and someone who can model the plate / key / cap in a 3D program to check for interference.

[ideus]

Interesting idea, however there is no way the cutouts for the switches remain the same as in a flat plate due to the curvature. Even small curvatures will alter them and how the switches keep in place.

[IvanIvanovich]

Kinesis managed it... perhaps some study on how they have done it... though they used injection molding plastic. I think the concept could be carried over to metal though, but everything would need to be super precise and have very tight tolerance.

[ideus]

Molds can be fine tuned to get exactly what you need in terms of final shape, dies for metal in the other hand face a great challenge with spring back effect of the steel. Otherwise a machined metal plate would cost a fortune to have them cutted out of a solid block.

[E TwentyNine]

Didn't someone try (or at least propose) this a while back?

If the curve is a problem for plate mounted switches, why do a smooth curve at all?

Do a "rough curve" were each row is flat using a sheetmetal bending brake.



Artist depiction of a three row keyboard done in this manner:

     |
 __/

[ideus]

Didn't someone try (or at least propose) this a while back?

If the curve is a problem for plate mounted switches, why do a smooth curve at all?

Do a "rough curve" were each row is flat using a sheetmetal bending brake.

Show Image

Artist depiction of a three row keyboard done in this manner:

     |
 __/

That is totally doable, the next problem then is the pcb, or maybe the board should be direct wired.

[0100010]

Didn't someone try (or at least propose) this a while back?

If the curve is a problem for plate mounted switches, why do a smooth curve at all?

Do a "rough curve" were each row is flat using a sheetmetal bending brake.

Show Image

Artist depiction of a three row keyboard done in this manner:

     |
 __/

Exactly.  Couldn't be a complete curve.  Would have to keep each row flat, with a bend in between each row.

[E TwentyNine]

Realized the stepped curve approach could be an issue for vertical keys.

If it were a smooth curve, would a cherry keyswitch mount ok with just the front and back of the switch in full contact with the plate?

[Findecanor]

The edges of the hole need to be straight also.

Another approach would be to make a plate for each row of switches. Then place them into a jig and spot-weld them together into a single solid piece. Alternatively, just attach them really well to the case/chassis.
If there is a vertical key then it could have its own flat plate-piece.

Anyway... I am not convinced that curved backplanes are better than flat from an ergonomic standpoint. With a curved backplane, the further a row of keys is from the wrist, the more the key-pressing motion would be pressing away rather than down/towards you.
I think curvature would make sense only on alphabetic rows below the home row, which on most keyboards is only one: the ZXCVB row. (inverted-T cursor keys not included, because you would move your hand to reach them)
BTW, when discussing the distance and spacing between keys when the keys are on a curved plate or dished keywell, I think that one should use as reference a point on the front on the top surface of the keycap when the key has reached its activation point.

[Photekq]

What would be the advantage of a plate like this? I just don't think there would be one. We would also need completely custom uniform profile keycaps for it.

[Findecanor]

We would also need completely custom uniform profile keycaps for it.

I suspect that someone has Signature Plastics' DSA profile in mind... which is quite common in group buys by now.

The biggest benefit would be that you could just move these caps around for any logical layout. No need to have different sets for QWERTY, Dvorak, Colemak, Neo etc.

[E TwentyNine]

I'm convinced curved plates are better than flat for one simple reason. Look at these images:



For the curved plate, the plane of the key surface is perpendicular to the travel of the key.  You press in the direction the key is moving.

For the straight plate, the plane of the surface is angled off the perpendicular.  You press in a way that's not in line with the key travel.

[Oobly]

Curved plates have a number of benefits and I like the idea a lot! If you look carefully at Cherry, DCS and OEM profile you can see that they will actually all work with curved plates without a problem. It would even make DSA feel more natural.

A plate bent between the switch holes should be doable and you don't have to separate the switches all that much. In fact, if you only space them a tiny bit further apart so that the bottom of the caps are at the correct distance, the keycap tops will be closer together, allowing less finger movement for typing 

If you want to use PCB's you can use one per row, but direct wiring would be easiest I reckon.

The two vertical keys can have their own little plate which is somehow then connected to the main plate.

I want a 60% like this with Cherry profile caps. No vertical keys to complicate things.

[jacobolus]

Anyway... I am not convinced that curved backplanes are better than flat from an ergonomic standpoint. With a curved backplane, the further a row of keys is from the wrist, the more the key-pressing motion would be pressing away rather than down/towards you.

I agree.

I'm convinced curved plates are better than flat for one simple reason.[...] For the curved plate, the plane of the key surface is perpendicular to the travel of the key.  You press in the direction the key is moving.

For the straight plate, the plane of the surface is angled off the perpendicular.  You press in a way that's not in line with the key travel.

This seems incorrect to me. When you move your finger to find the new key, you’re extending your two distal joints outward/upward, but when you actually press the key down, the main motion is still to flex the whole finger at the first knuckle (metacarpal-phalangeal joint), and the direction of force is fairly close to straight down, not outward. (At least, that’s what it seems to me from observing my own hands’ movement.)

This is one of the things I find to be problematic on the Kinesis Advantage: many of the keys travel in an axis that is not the same as the axis of finger pressing, and as a result they are less uncomfortable to press than they would be if rotated to be more straight up and down. The Maltron has an easier time getting this closer to right, because it uses hand wiring instead of a curved PCB.

I do agree though that there might be an advantage in raising the further-back keys up higher than current keycap profiles permit. So perhaps making "steps" of sheet metal (Or CNC'd metal/wood/plastic) could be nicer to type on than current flat designs. Also, the proximal keys (bottom row or two, nearest the body) could probably be angled a little bit, as I think they tend to get pressed a bit toward the body as well as down.

In general, my expectation is that you get a much bigger ergonomic advantage from splitting the two hands apart, tenting them, and putting thumb keys in a comfortable spot, than you get from curving the plate the finger keys sit on. If you feel a pressing need for height differences in the finger section, I’d recommend switching to a column-staggered layout and adjusting the height between columns pressed by different fingers, e.g. raising the pinky keys up a bit. (But don’t try to accomplish this with a curved plate/pcb, or you’ll get the same problem again, where the pinky/index finger keys are pressed by unnaturally pushing sideways.)

[0100010]

So for a 103 key ANSI layout, would need to cut the plate like this :



Where all the single row keys are on the main plate, and the numpad + and Enter keys on on their own smaller plate.

Use a brake to apply a ~3 degree bend between each row, and probably around the same amount between the two keys on the smaller plate.


As far as keypress direction - it makes sense to me that you would want to push the cap in the direction of travel of the switch.  Model Ms are like this, and many folks seem to like them.

[Michael]

What would be the advantage of a plate like this? I just don't think there would be one. We would also need completely custom uniform profile keycaps for it.

There was a flat buckling spring IBM kb that used ergo caps (pbt, dye subbed), the same used on the ergo version of the IBM. Don't know how hard it would be to source those - perhaps unicomp might be able to help in that area

[RabRhee]

Take it to the next stage, where the plate is 6 strips, with an adjustable lockable hinge between rows so you can vary the curve row to row. Note. the lock will need to be pretty strong or a heavy hitter will flatten the board

[jacobolus]

As far as keypress direction - it makes sense to me that you would want to push the cap in the direction of travel of the switch.

That’s correct. But in general, that direction is pretty close to straight down.

Model Ms are like this, and many folks seem to like them.

The angle of Model M switches is suboptimal. Just like the layout. But it works okay, because the angle is actually quite slight.

A Model M where you tilted the near side of the keyboard upward by about 5 degrees would probably be a bit better to type on.

[0100010]

The angle of Model M switches is suboptimal. Just like the layout. But it works okay, because the angle is actually quite slight.

A Model M where you tilted the near side of the keyboard upward by about 5 degrees would probably be a bit better to type on.

Can solve for the angle of the bottom row with the case for a plate like this.  Instead of having the bottom row be parallel to the desk, make the 2nd to bottom row parallel.  That would put the bottom row at a reverse 3 (or 4) degree angle.

Example :

[Melvang]

I have actually been sourcing some parts for something exactly like this.  I have a QFR plate and supposed to be getting a terminal board plate from an M from dorkvader (he still needs to reply to my pm).  First I am just going to drill and tap some holes in the QFR plate and attempt to pull that plate to the M plate.  The cherry switches only clip to the plate on the top and bottom of the switch, aside from numpad + and Enter.  This is why I chose a TKL layout.  If this doesn't work I will try a bend between the switch rows. 

PCB won't be a problem with the Enablers. 

[bueller]

PCB won't be a problem with the Enablers.

Was thinking the same, or even doing individual PCB's for each row.

[Melvang]

PCB won't be a problem with the Enablers.

Was thinking the same, or even doing individual PCB's for each row.

That would work as well but won't be as flexible on layout.  Granted you are still limited by the plate but that's the cheaper part between the two.

[jdcarpe]

So, we need a plate set like this. Then you can spot weld the plates together. This is for a TKL, so no vertical key issues.

[Melvang]

So, we need a plate set like this. Then you can spot weld the plates together. This is for a TKL, so no vertical key issues.

Show Image

I would vote small stitches with TIG vs spot welds.  The two plates would have to touch to spot weld and I am not sure how well that works right on the edge of the part.

[mkawa]

curved PCBs are _really_ expensive. you have to lay down the fiberglass mat in a curved mold and then shoot the resin binder to make curved FR4. it's a huge tooling investment just to make pcbs. jd's solution is _much_ better

[Melvang]

curved PCBs are _really_ expensive. you have to lay down the fiberglass mat in a curved mold and then shoot the resin binder to make curved FR4. it's a huge tooling investment just to make pcbs. jd's solution is _much_ better

JD was just talking about the plate.  But that idea would work for the PCB as well, single rows tied together with hookup wire.  I was going to use enablers for my experiment.

Think I could get 2 40% boards out of one M plate going this route?

[jdcarpe]

So, we need a plate set like this. Then you can spot weld the plates together. This is for a TKL, so no vertical key issues.

Show Image

I would vote small stitches with TIG vs spot welds.  The two plates would have to touch to spot weld and I am not sure how well that works right on the edge of the part.

I defer to the expert on this one. You know WAY more about welding et cetera than I.

[Melvang]

So, we need a plate set like this. Then you can spot weld the plates together. This is for a TKL, so no vertical key issues.

Show Image

I would vote small stitches with TIG vs spot welds.  The two plates would have to touch to spot weld and I am not sure how well that works right on the edge of the part.

I defer to the expert on this one. You know WAY more about welding et cetera than I.

I will be attempting this.  Though not sure exactly what method I am going to use for cutting the plate.  I have some options available but not sure yet.  At least the QFR and the M plates are both carbon steel so that makes the welding much easier.  Then the next problem comes with getting proper holes in the M plate without removing to much material and losing the support from that. 

I wonder if mkawas WJ guy has the capability to cut square holes out of an already formed M plate?  Or if it would be cheaper to cut holes then form the plate.  That should work as well cause the sides of the hole don't clip to the switch at all.

[CPTBadAss]

What if you extended those plate edges to have lips so you could form a proper lap joint to weld instead of tacking it?

[Melvang]

What if you extended those plate edges to have lips so you could form a proper butt joint to weld instead of tacking it?

Still wouldn't have any support in the middle of the row.  Besides, lap joint is stronger than a butt joint.

The only way I can see not having to cut holes in the M plate is to space the TKL plate/row plates off the M plate with standoffs.  This might prove a labor of love because of the limited space between rows/columns, and the fact that spacing it out like that would result in a smaller radius for the TKL plate.

If you can't tell, I have been thinking about this idea quite a bit recently.

[CPTBadAss]

What if the plates are designed with removable bosses which can hold standoffs? Think of how the Model M barrels are removable. Instead of holding a switch, it accepts a standoff.

And then lap joint it 'cause I was silly and tried to sneak butt joints past you.

[Melvang]

What if the plates are designed with removable bosses which can hold standoffs? Think of how the Model M barrels are removable. Instead of holding a switch, it accepts a standoff.

And then lap joint it 'cause I was silly and tried to sneak butt joints past you.

So pressing/forming a plate with square raised holes to accept the switch for a one piece plate, or am I thinking the wrong direction?  I am a very visual person.  Even napkin CAD works for me.

I am thinking double male ended standoffs.  One end threads into the Cherry plate, the other end is through hole in the M plate for a nut.  Or same thing but with a nut on top and bottom of the M plate.  With this technically you would have adjustable height and pseudo-adjustable radius.

[CPTBadAss]

I can't sketch at work. I will post something later.

Edit: Someone PM me a reminder if they would be so kind. I am insanely forgetful.

[MOZ]

Subbed.

[Melvang]

I can't sketch at work. I will post something later.

Edit: Someone PM me a reminder if they would be so kind. I am insanely forgetful.

See edit.

Napkin cad on break?

[CPTBadAss]

What's a break? And even if I sketched something, I can get fired for taking pictures on my phone here.

[Melvang]

What's a break? And even if I sketched something, I can get fired for taking pictures on my phone here.

Understandable, in your field they would probably stick to that as well.  Couple guys got caught taking pics on the melt line at the foundry.  They about got walked to the gate on the spot.

[CPTBadAss]

I was given a very stern talk the last time I took pictures at work. Not risking it again.

[Melvang]

I was given a very stern talk the last time I took pictures at work. Not risking it again.

Yeah, the only job I could see being worth it would be fast food.

[0100010]

So, we need a plate set like this. Then you can spot weld the plates together. This is for a TKL, so no vertical key issues.

Show Image

Would probably have to prototype both a plate that was originally flat, then bent along the rows and a plate that was cut rows that were stitch welded together.

Would imagine it would be easier to get a consistent bend angle between the rows using a brake (with likely more chance of hole distortion), where a welded plate would be hard to get a consistent angle without a jig (but also very little chance of hole distortion).

[CPTBadAss]

That's why i was suggesting extra material. You could get the bend in those plate strips and then if would help with some of the distortion.

Edit: Now that I think about it...maybe that's why we want to tack. I don't have *that* much personal experience with welding so maybe Melvang can chime in.

[jdcarpe]

So, we need a plate set like this. Then you can spot weld the plates together. This is for a TKL, so no vertical key issues.

Show Image

Would probably have to prototype both a plate that was originally flat, then bent along the rows and a plate that was cut rows that were stitch welded together.

Would imagine it would be easier to get a consistent bend angle between the rows using a brake (with likely more chance of hole distortion), where a welded plate would be hard to get a consistent angle without a jig (but also very little chance of hole distortion).

My sheet metal guy advised me that in order to get a good bend without distorting the holes, there needs to be at least 0.25" clearance between the edge of the hole and the bend. That would mean 0.5" between rows, at minimum. :(

[CPTBadAss]

What if you bent the plate so that it formed a lap joint underneath? Sort of like this:

------
       | ------------------
       ----------

Lines represent 2 plates. I'll make a sketch for this later too.

You bend the plates to get the angle and then make that joint to bring the rows closer together before welding the sheets together.

[Melvang]

That's why i was suggesting extra material. You could get the bend in those plate strips and then if would help with some of the distortion.

Edit: Now that I think about it...maybe that's why we want to tack. I don't have *that* much personal experience with welding so maybe Melvang can chime in.

I think someone needs to do some extensive math to see how much the caps will move together when bent to match an M plate.  With what I was thinking if you placed an M plate next to the bent plate the bends and the ends would match up.  I know distance between caps at the bottom of the skirt on DCS caps with .75" (19.05mm) spacing is .035".  This would end up being a weekend project as I haven't delved that far into geometry and the bit I have had was in high school (15 year reunion next year).  I am pretty sure I could get the answer, but someone good with cad/modeling could probably come up with a quicker answer.

For what it's worth I also know DSA caps are a tighter fit between caps but I don't have numbers for those.  DCS caps single unit caps are .715"x.715" according to SP.  Though I don't know how far from the plate the bottom of the skirt is.

Coming to this answer would tell us if/how much extra material we would need between rows.  Though I think for the sake of modeling I think the best thing to do is just to get an M plate and a stock plate and start cutting and welding.  I can probably handle this part but I don't have a spare M plate.  For testing purposes I can just cut it apart in my garage and stick together with some poster tack.

[CPTBadAss]

I got you on the M. PM or text me and we can work it out.

[E TwentyNine]

My sheet metal guy advised me that in order to get a good bend without distorting the holes, there needs to be at least 0.25" clearance between the edge of the hole and the bend. That would mean 0.5" between rows, at minimum.

How realistic is it to cut the holes after the bends are done?

[jdcarpe]

My sheet metal guy advised me that in order to get a good bend without distorting the holes, there needs to be at least 0.25" clearance between the edge of the hole and the bend. That would mean 0.5" between rows, at minimum.

How realistic is it to cut the holes after the bends are done?

Probably not at all. The laser needs clearance to move, and basically wants a flat surface to cut. Plus clamping the plate down to get each row flat with the table for cutting would be nigh impossible as well.

[Melvang]

My sheet metal guy advised me that in order to get a good bend without distorting the holes, there needs to be at least 0.25" clearance between the edge of the hole and the bend. That would mean 0.5" between rows, at minimum.

How realistic is it to cut the holes after the bends are done?

Probably not at all. The laser needs clearance to move, and basically wants a flat surface to cut. Plus clamping the plate down to get each row flat with the table for cutting would be nigh impossible as well.

5 axis (i think) waterjet would be the key here.