Silicone Keypad Design Problems
Electricity and water is the issue. The tin in the solder plating will grow a crystalline structure and form an oxide that doesn't conduct very well. Spent many months in the 1980s solving this problem and the bottom line is use gold plate. Don't be cheap on this. The company I worked for at the time sued the supplier a lot of money for their incompetence and they were big in the industry at that time.
If you can't seal it (and clearly you can't because you can clean the contacts) then water will get in. It's inevitable.
If you can't afford decently thick* gold plating, I suggest you specify the PCB with printed carbon conductive ink. HASL or tin plate will not be reliable over the long term and what you have experienced is to be expected. Nickel would be better, but still not great.
You will find that conductive ink is standard in most consumer remote controls, and something like that is what is on the other side of the contact (a carbon-loaded elastomer pill).
If you can't find PCB makers willing to do that in your quantities, get the gold (it should always be supplied plated over a nickel barrier layer) and be done with it.
*ENIG (electroless nickel immersion gold) is not really to be strongly recommended for keypad use- too thin, only a few microns thick.
Hard (electrolytic) gold over nickel is literally the gold standard for contact surfaces. Unfortunately, the gold has a negative effect on solder connections (there are typically too few gold atoms in ENIG to seriously embrittle the joints for most applications) so it should be limited to areas that are not soldered or it must be removed later (as is detailed in the IPC J-STD) for high reliability applications.
J STD-001 Revision “F” now states: (note that the new wording/changes are highlighted below) 4.5.1 Gold Removal
Gold removal is performed to reduce the risk of failure associated with embrittled solder. Gold embrittlement is not a visually inspectable anomaly. In cases where analysis has determined there is a gold embrittlement condition, the gold embrittlement shall be considered a defect, see IPC-HDBK-001 or IPC-AJ-820 handbook for guidance. Except as noted above, gold shall be removed:
a. From at least 95% of the surfaces to be soldered of the through-hole component leads with >2.54 μm [100 μin] gold thickness and all through-hole leads that will be hand soldered regardless of gold thickness. b. From 95% of all surfaces to be soldered of surface mount components regardless of gold thickness. c. From the surfaces to be soldered of solder terminals plated with >2.54 μm [100 μin] gold thickness and from all solder cup terminals, regardless of gold thickness.
Late Edition
Other new process are now available.
- Nanofics technology uses a low pressure, dry plasma system to deposit fluoropolymer nanocoatings that provide permanent hydrophobicity and/or oleophobicity. The system is inherently “green,” and does not generate chemical waste; coatings are PFOA and PFOS-free. ( Used to make cellphones waterproof unlike every Asus and iOS product ( That I have owned and seen fail from moisture) that rust with moisture from lead-free corrosive no clean flux)
New Chemistry Deposits 4-8 µin
Uyemura has introduced a reduction-assisted immersion bath for board customers who demand an immersion gold deposit above the standard 1-2 µin on ENEPIG. Called TWX-40, this is a mixed reaction bath – an elite hybrid – that delivers both immersion and autocatalytic (electroless) modes of deposition.
TWX-40 is a proven alternative to other attempts to achieve heavier gold deposits on ENEPIG, (electrless Ni then electroless Palladium then immersion Gold) Cu > Ni > Pa > Au
It may come down to who has the better process controls.
Carbon Ink with short shelf life or ENIG from porosity and even distribution of paladium catalyst or C sense with finger noise .
*One of the biggest Process Improvements was a saturation gold density limit with monopolarity pulse electroless plating. They now use a specific burst profile with periods of reverse polarity pulses, repeated in profile for desired characteristics and is faster and cheaper than traditional ENIG or EP
This also provides finer gold deposits and higher density for less porosity.
In any case CPk or 3 sigma analysis on SNR ratio with accelerated humidity heat cycles are suggested. Min SNR of 10 worst case without skew meaning threshold under all conditions. Robotic fingers however do not replicate human touch with side swipe. e.g. Bosch have terrible low sensitivity touch sensors in some appliances.(due to finger capacitance variations and internal settings too high)
Another is a thin plastic film between metallized conductors on urethane membrane and the electronics only measure the capacitance change.
Anecdote
( reminds me of an auto car wash with digits for wash code in Toronto and the membrane buttons were gouged out with keys and pens because frantic users wanted to get their car washed without delays.)
Carbon ink works but softer and reliability depends on user excess abrasive pressure which can also be a reliability problem. My old keyfob and old car garage door opener used this and it is wearing out now.
Details
Electroless Nickel Immersion Gold (ENIG) is a type of surface plating used for PCB with a thin layer of immersion gold, which protects the nickel from oxidation using a Palladium catalyst while the Nickel is electroless plating over copper.
ENIG has several advantages over more conventional (and cheaper) surface platings such as HASL (solder), including excellent surface planarity , good oxidation resistance, and good usability for untreated contact surfaces such as membrane switches and contact points.
IPC Standard IPC-4552 covers the quality and other aspects of ENIG finish on printed circuit boards. IPC-7095 covers some "black pad" related features such as so called mud crack appearance and nickel protrusion spike
Ref
Other
The other requirement of membrane switches is to have at least 15mm air clearance to human finger or else a plastic insulation layer with 15kV breakdown protection against ESD..