The bottom line on rubber buttons is that they are done for a few reasons, very few of which are applicable or necessary to hobbyists.
- Low cost for high volume products, both in part cost and assembly costs
- Surface/splash water resistance
- High cycle count applications (elastomers can be engineered to be amazingly durable at the cost of actuation force and tactile feedback)
- Applications such as medical that require a sealed keypad that can be cleaned and sterilized
If someone is out there trying to build an elastomeric keypad with carbon pucks for switching because, well, that's what the big kids do, my advice is: don't. Rubber keypads with carbon pucks are largely used because they are cheap, not because they are necessarily that good. Anyone that's had a TV remote where the rubber buttons became balky can attest to this. A plastic key over a metal snap dome or a low-profile microswitch will almost invariably be cheaper, easier, more reliable, easier to repair, and have better tactile feedback. In fact, my current cable remote control has gone over to this method: rubber over metallic switches.
If you're building a proper keyboard, my advice will change from the above.
An elastomeric key that functions like you're used to with your TV remote is actually a fairly sophisticated piece of engineering. The portion of the switch that flexes in order to create the snap-like tactile feedback is paper thin has a very specific profile and angle. Every, single key must have this delicate, highly-engineered, membrane surrounding the thick, rubber key pad or the key will seem stiff, mushy, lifeless, or worse, fail in the closed position. Additionally, each key needs a vent to the outside world, or you will simply have a suction cup. These vents are very shallow in order to not compromise the structure of the key. Another parameter is the durometer of the rubber. Softer durometers give better feel, but at the expense of life and reliability.
The contacts are also interesting. Everyone seems to know about the carbon pucks. What gets left out is the carbon silk screening over the top of the switch pads. This is essential for a switch that will work over the long term. Going without can work for a while, but carbon contacting carbon is the commercial solution.
Once you get past the above, then you get to worry about how the switches interact with whatever bezel, if any, covers the switch matrix. This gets interesting quickly. Despite appearances, there's a lot to the clearances and angles in order to have a key that consistently actuates when pressed and does not stick in the down position by becoming jammed on a sidewall on the return.
I see mentions of super-cheap custom key matrixes. Granted, none of this is remotely rocket science, but I can virtually guarantee that you cannot get anything decent at the prices quoted after all of the time and expenses are tallied. If these shops were any good at all, they would not be selling $250 tooling. You might get something in the shape you want, but it's just going to be a shitty key. They aren't charging enough to make anything good. There's an enormous amount of know-how and trial and error that goes into these things, and you need to be sure that you really, really want to go down this road before embarking upon such a venture.
And the reality is, that rubber matrixes aren't so attractive that I would forgo other options, particularly on low-volume devices. I would strongly suggest other alternatives such as a polycarbonate membrane switch, plastic key over a switch or snap dome. The road to Rome is far easier.
What's great about today is that some of the very low cost tactile switches coming out of China are FANTASTIC. They feel good and some of them put out close to perfect square waves. Obviously, you'll need a debounce routine, but the old cost/quality arguments against microswitches just aren't there like the bad, old days.
The rubber key matrix certainly has its legitimate place and will for a long time to come, but make no mistake, it's not a $250 slam-dunk to implement, and there are far more attractive alternatives for prototypes and low-volume. People have been trying to cost reduce key matrix tooling and production for decades. If RTV and cast urethane worked, it would have been done loooong ago. But it doesn't. What is being suggested is hugely time intensive and the results will be completely unsatisfactory, unless one just has incredibly low expectations.
Buying a pre-made matrix off-the-shelf like Frank is going to, is totally legit, and recommended if that works for you.