Un-flaking a super-touchy laptop key

  I bought a pair of laptops with the general idea of upgraded computing platforms for both me and an aging parent; our old Macbooks were no longer quite up to speed with what the modern internet could throw at them.  Rather than go the Mac route again, I wanted to move to one of the newer Linux distros and accept the challenge of making it as easy-to-use as I could.  There was no chance in hell I was going to saddle either of my intended recipients with the preinstalled Windows-10, and it always brings a little pleasure to summarily overwrite that garbage with something real.  As a platform I settled on Lubuntu, a "lightweight" version of the popular Ubuntu desktop environment.  [Which influences my choice of color theme for this page...]  It's not so lightweight in geek-level reality; it is burdened by the confusing "systemd" boot and configuration setup, which for a traditional Unix admin is a huge hairball learning curve and many "WTF were they thinking" moments.  [Read my rant on that topic, and for the record Samsung blew me off and refused to answer my honest question about flash cells.]  Fortunately I had a little prior experience with systemd [mostly figuring out how disable stuff], and once that's all done and a system is running the way we want, very little of that is visible to the ordinary user.  Set up a few clicky-buttons to launch common apps, and away we go.

Laptop opened up to replace NVMe module For the unit I'd keep I had much more geeky intentions, so to make it a more significant upgrade from the Mac it needed a bigger and more reliable NVMe storage module.  The stock installed one turned out to be a Samsung V-NAND type anyway, but in the more consumer-grade EVO class.  Here I would double the size and go up a performance tier to the "pro" series, and for my very first foray into the world of M.2 and NVMe this was already a significant adventure.  While there's no convenient hatch for field-swapping storage in these units, at least it's a relatively few screws and some gentle prying to get the whole bottom cover off and expose everything.  [Lenovo Thinkpad T series, by the way, sort of following the tradition from my geriatric but beloved and still quite functional 600x.]
This is immediately pre-swap: the 970-PRO is sitting in the carrier from a small external USB enclosure I also ordered.  The laptop's original module wound up in that as a plug-in drive.  Most NVMe modules run hot when in their active power states, and need to be heatsunk to a large metal surface on top with a thermally-conductive pad.  The laptop's layout can clearly also accept an ordinary 2.5" SATA drive in the same spot, but here the metal shield that would go around that is used as a generous heatsink for a NVMe installation instead.  The assembly tucks under the right-hand keyboard palm pad, thus providing immediate human feedback as to how hot the drive might be running.

  What I didn't count on was running into hardware problems on brand-new gear, especially when the symptom almost looked like software driver issues at first.  And it showed up on the *other* machine of the pair, i.e. not the one I'd just taken apart.  As I started working for real on both of them, the non-upgraded one seemed to occasionally issue additional keystrokes when some low digits keys were typed, especially the '1', and my first thought was that the Linux kernel keyboard mapping or timing was screwed up somehow.  Later I did more testing and discovered that the barest whisper of touch on the F1 key would cause it to assert, sending "<esc>OP" into a typical terminal window.  It was sensitive enough that typing a '1' just below it would usually trigger the false F1 event as well, just through vibration or slight flexing of the keyboard frame underneath -- if I tapped my '1' very lightly I could sometimes avoid it, but it was a pain in the ass and only likely to get worse over time.

The meta-problem was that it was happening on the machine I intended to send out, and had already spent *hours* of work toward setting it up.  Meaning that sending it back wholesale for warranty replacement was not in the cards, at least not without removing or zeroing its storage -- I needed to simply try and fix this. 


[Images are linked to larger copies.]
Keytop removed, pantograph out It's generally easy enough to pop keytops off, but the caveat on these is that they clip down at the upper edge and the lower one is actually hooked under tabs on the little "pantograph" stabilizer underneath.  Going carefully with a magnifier and flashlight shows the right way to do this without breaking anything.  Next challenge is to get the pantograph out, which involves spreading the edge rails of its outer piece enough to disengage from the tiny pins on the inner part.  Tweezers or small needle-nose pliers and a couple of angled probes help [I have some old dental picks].  We obviously make sure to record the proper orientation of everything.

The laptop goes under the microscope I needed a very close look at all this, so the whole shebang went under the dissecting scope and I had my micro-surgery tools ready.  My first suspicion was stray gunk under the membrane, but very close examination showed the space under the nipple to be clear and nothing actually looked obviously wrong.  The lower edges of the dome structure weren't totally sealed; I found three small but intentional holes around its outer perimeter, presumably for air-pressure relief, through which a small tool could easily go in and scrape around to make sure nothing extra was between the nipple and metallic switch pad areas that actually make the connection.

  I powered up the machine just into BIOS setup -- easy enough to type F1 to get into that, huh?  And a further F1 press jumped to its help screen, so I had an easy way to test for an asserted keypress.  Eventually I noticed that the nipple/switch structure was most sensitive toward the front edge of the laptop, i.e. tilting the dome toward the rest of the keyboard or lightly touching the base near where the circuit traces come in would trigger the problem.  When I lifted the whole section of membrane up slightly, the problem wasn't as pronounced, so perhaps pressure from the frame plate underneath was unduly pressing the switch pads together.  There wasn't any obvious dirt in the way, it was just some sort of internal defect.

The whole circuit matrix is constructed of layers of flexible plastic membrane with the traces basically printed onto them, all adhered together.  The design relies on the layers maintaining a small gap between the printed pads at rest, and a keypress flexes the top down to make contact.  Somehow these two pads were already sitting too close together, where the slightest disturbance basically made them touch.  Longish story of farting around cut shorter, I decided that the two metallic pieces of the switch needed to sit farther apart where the feed wires arrive into it.  Now, how to go about doing that??


Splitting the keyboard membrane Fortunately, this design isn't a big contiguous sheet of plastic as many others are, and I was able to carefully raise the area and split the layers apart a little.  With the layers held open even a little bit, the sensitivity was gone.  Simply letting them re-settle back onto the disturbed glue still wasn't enough to keep it that way, which I was kind of hoping for, so it needed something extra.

The fix: a tiny slip of paper The eventual answer was simply this: a tiny slip of paper inserted between the layers just ahead of the switch proper, the whole assembly stuck back down in place, and the excess paper cut away.  That thickness is enough to keep the switch pads apart but still allow a proper keypress to connect them.

The biggest pain in the ass of the whole operation was getting the little plastic pantograph back in, which needs about three hands and pairs of tweezers at once.  But eventually it all went back together, and the stress relief when the key-top finally clicked back down in place was profound.


Classic poster of duck with mallet, hit any key F1 now behaved like any other key, and I could feel much more confident about placing the machine in the hands of a relative computerphobe.  No idea exactly what was wrong in the switch, or if it's even a common problem.  Perhaps some shop or assembly tech had followed the advice of the classic office poster from the early Nineties, where the key in question was often "F1" rather than "any".  I was just glad I could deal with it on my own without agonizing go-rounds of negotiation and shipping with the original seller.

_H*   200720