Gents, in a series circuit, where there are two resistors R1 (the pull-up), and R2 (the resistance of the contact at the plumb bob from bi-metal contact), the higher the pull up resistance (R1), the greater the voltage drop will be measured (whatever the current) across that part of the series network. So in the conductance case, you'd be right, with all that voltage across R1 (especially if we pull it higher), the lower the signal would drop as measured across the contact (R2). This would be good, as you guys describe.
But we are not seeing "contact", and the reason isn't the voltage, it is the *width* of the voltage pulse (zacaj' picture above is, after taking away the double pulsing, pretty generous. The normal pulse is 15 - 40 mSecs active-low depending on the board) . We are trying to increase *reactance*, not conductance - the inverse of capacitance across that contact. Which means we hope to pump current across those contacts to overcome whatever the RC value of that contact is (lower the pull-up resistor). I can't change the filtering capacitance on the input circuit on the board, or when you press the start button, you'd register more than one "game start" request occasionally. So I have to change the 'Q' response of the contact during a pulse (skip the rest of the parts of that equation that I cannot change, like chokes and crap) - to do that, I have to pump more current: i.e. lower the resistance of the pull-up, and hope that the resistance of the plumb-bob contact is still minimal when compared to a 1K pull-up. When you think about it, at what point would it make a difference if the pull up is 10K or 1K - probably right around 100 Ohms contact resistance value. I don't think that this is the case, so I doubt that we care about the resistance of the pull-up.
Having said all that - it's a crap shoot. At design time, you start somewhere, put a scope on the signal (with capture), set your scan frequency, and see what the return signal looks like as we tweak the return capacitance / resistance values, coupled with the strobe signal pull up resistance. Fun huh?
Now, to answer the question: on the original board, we did logic debouncing of a contact - in other words, a valid switch is identified after 3 active low pulses are received sequentially on the return line from the same strobe line at the 8279. That is not normal for a pinball CPU, but at the time, I was into acoustics, and a purist, this was how we did things to be absolutely sure that transient signal spikes from the rest of the game were not being seen on the return circuits showing false signals. In 1995 it worked great! Even for the plumb bob (or so I thought then with the game I had) It was partly this circuit design that allowed us to really improve spinner target response. We could detect spinner closure 2 - 3 times the amount that the original board could, and this was one of the odd goals we had (remember the reactance spiel). With the new board, we tried conditioning the return pulse with a more complex circuit (check the schematics between the 2 boards), and do away with switch debounce. Embarrassingly, that worked just as well, and the response seems about the same. It cost more, but the return circuits were pretty much bullet-proof - you could short a coil voltage to the returns, and not blink (though, that would probably take out the strobes).
Now, how fast are they strobed? I can look it up, but remember we are strobing the displays at 2X the speed of the switches. We need to fit 8 strobe lines into one full strobe "cycle". Whenever I change the active low pulse width, I also wind up changing the width of the full pulse cycle on the displays. I think the default width is 15 mSecs as I mentioned above - the original board used closer to 40 mSecs active low, which will definitely help improve switch detection of the plumb bob, but then you've lost your spinner response, (and overall playfield response feel).
Now, you can change this on the new board: set the options to EXPERT mode and tweak the switch strobe speed. It shouldn't affect the game - though it might make the displays a little less crisp. Don't go past 40mSecs. This will help the plumb bob response, by reading the signal, hopefully after the reactance curve falls to the acceptable level, and won't affect the front door switches.
In the end, everything is a trade-off. Electronically, you get fast game play, or a good plumb bob. If you want to run a System 1 tournament, clean the bob, and the brass contact ring, you won't be disappointed, if that is still not great, set the switch strobe speed higher, and put the game into Tournament mode. Sorry, this isn't an option on the older boards.
-Ace