Happy Springs: Using 3rd party mylar on bucking springs

[ricercar]

I bid on an eBay HHKB Lite 2 thinking it was an expensive model. Fortunately the auction was not obscene; I paid only $25. After learning this was a rubber dome board, I was going to populate it with Cherry browns.

However, when I opened up the HHKB, the domes and mylar sheeting were backed up by a metal plate. Hmmmmm. I've wanted to make a buckling spring keyboard smaller than my Space Saving mini. And lookie here. The HHKB mylar spacing is just the same as the Model M spring frame.



I have a pair of Model Ms that I've been trying to restore with a bolt mod. Unfortunately the spring frame is cracking, almost all the the key rows are separating from one another. The lower half is cracked entirely through, so that was my experimental platform.

The proof of concept was successful. All of the keys in the following picture operated. The others I figured were mis-aligned spring hammers.



I cut the rubber sheet from one of my Model Ms.



The following image shows some separated key rows from my Model Ms. The row-by-row horizontal spacing was near identical to the same-row mylar pads. Row 1 spring frame matched the spacing of row 1 HHKB mylar.

Vertical spacing between rows was slightly greater on the mylar than the spring frame. Unfortunately there aren't sufficient holes in the metal plate to mount row 3 separate from row 4. I decided the initial experiment justified the sacrifice of my second Model M spring Frame, which had only one crack, between row 2 and 3, between QWERTY and ASDFG.



With this amount of progress, I hooked up the controller to test. There was significant delay between the mechanical and audible click, and the electrical key registration. This was really unbearable.



Another issue was the discrepancy for the left shift key, right shift key, and the FN key on row 4. They would need separated spring holders to line up with the HHKB mylar.



With the lag between key press and key registration so huge, I removed the rubber sheet to see if it made any difference. It didn't. So I left the project unfinished, without fully resolving the arrow cluster, or function keys.

Here's the most advanced stage of the experiment.



The HHKB is now restored to its ugly bestial norm.

[ricercar]

****ing editor dropped 30-some minutes of editing. I'm too pissed to retype.

Summary: don't do this. Key response was too slow for humans to type at real-world speeds. Details to come with restoration of my inner peace.

[ricercar]

Article re-edited with smaller and more numerous photos. I still hate this editor passionately.

[quadibloc]

I am just truly astonished by this mod. A buckling spring HHKB. But why not, as a Model M mechanism uses the buckling spring on plastic contact sheets not that dissimilar from what you might find in a rubber dome keyboard... although the cursor keys, being small, would be tricky.

[itlnstln]

Mmmm... BS HHKB.  That would be awesome.  Lowpoly, watch out.  You're getting some competition over here.

[ricercar]

This occurred to me as I was posting the photos the second time. The hammer impact point isn't centered on the spring cylinder. Therefore eyeballing the Mylar dot contacts into the cylinder centers was not the right thing to have done.

Hypothesis one:
The hammer impact point was not aligned properly on the Mylar. If it was off-target slightly, A longer impact time creates an expanding area of depressed Mylar. Thus a longer duration key press could allow for an expanding area of pressed-together Mylar, with delayed electrical contact when the depressed area finally reaches the real contacts. This could cause milliseconds of delay with every keypress.

I'll test this hypothesis later this week. Unfortunately each such experiment requires new/widened mounting holes, so there's a limit to the number of locations I can try with one spring holder section. Unless I want to sacrifice a third Model M (nope). Well, hmm. well, I could instead get bigger bolts, up to the size of the holes in the metal plates...

[ricercar]

Yeah pressing the Mylar with my fingers was no problem. During the assembly, I was always making unintended electrical contact on the areas without buckling springs. Whups, I guess I was experimenting on a live subject. Bad human.

It's not the electrons that are delayed, it's the physical Mylar sheets. Pressing a key with more force did seem to help reduce the latency for some keys.

[itlnstln]

Well, first just try jabbing those dots with everything plugged in but no hammers to eliminate something else going on.  

Electrons go pretty damn fast on that short a distance.

Show Image

Drift velocity equation (don't ask me what values to plug in - I just copied and pasted in an effort to look smart).

I like the stealth effect of this diagram in the dark theme.  Well done, ripster.

[ch_123]

On a related note, I tried a Model F buckling spring with a Topre HHKB and no go. In fact, I could have sworn that I got a shock when I touched the spring...

[JBert]

In that case, the traces on the PCB on which the spring rests might not be enough of a surface to measure capacitance, hence the hammer of model F might not be detected.

Couldn't you place a bit of tinfoil on the outer contact but with a hole for the central contact? This way, the hammer might be sensed once you place it on top of it.

EDIT: Oh, and it could be that you felt a shock. The "capacitors" need to be charged after all.

[sethstorm]

I am just truly astonished by this mod. A buckling spring HHKB. But why not, as a Model M mechanism uses the buckling spring on plastic contact sheets not that dissimilar from what you might find in a rubber dome keyboard... although the cursor keys, being small, would be tricky.

Couldn't you just go the easier route of grabbing a Wheelwriter board, and use the appropriate controller?  I'd think that there are ones that are M-compatible but at the same time as minimal as the HHKB.  You'd have enough keys on a WW3/5 to approximate one.  That, and if I can get the matrix to one, I can see where the exact scancodes come out, providing for the possibility of doing it through USB as well.

See this for a reference of the wheelwriter boards:
http://geekhack.org/showpost.php?p=178367&postcount=27

[ricercar]

Experiment 2

EXECUTIVE SUMMARY
The keystroke latency appears to have been solved for a single row by mounting the spring holder approximately 2 mm higher.

DESCRIPTION
What this means is that I re-drilled the hole 2 mm higher for the springholder of rows 1234567890 and qwertyuiop.

Subsequently, the letter row responded quite well, with keys TAB through ] registering electrically 100% of the tested presses at the same time as the physical & audio clicking. UNFORTUNATELY this 2 mm difference totally disabled the number row. None of the keys registered at all, ESC through backtick. Delete/Backspace on the letter row also was non-functional.

Testing with one single key, the spacebar (edited, was erroneously "backspace"), also confirmed that the latency between key click and electrical registration can be acceptable. Despite the complications of the stabilizer, I found a location at which the space bar clicked and electrically closed always at the same time, and never clicked without an electrical contact.

Unfortunately this means each row must be separated and mounted individually to assure the buckling spring hammers align to the Mylar contacts. I'm not pursuing this mod, because this would require drilling at least 6 additional holes in the steel plate at locations that don't affect the traces in the Mylar sheeting. Such effort exceeds my ambition at this time.

[itlnstln]

Ah, yes, the "Executive Summary."  Because decision-makers want as little detail as possible in order to make bad decisions on a company's direction.  I am very bad at writing executive summaries.  I like to make informed decisions.