Ed explains it pretty wel in the C20+C21 topic. Between tantallum and ceramics I prefer ceramics personally.
Quoted from G-P-E:
Regardless of what was used before - there are right and wrong capacitors to use at C20 and C21.
Both of these two are part of the 74HC123 RC timing constant so maintaining some a somewhat consistent and proper capacitance value is important. Too low of value and the machine will reset when it isn't supposed to.
A ceramic capacitor most definitely CAN be used in this position. These caps typically have +/-20% tolerance - not real good for timing circuits. You can get these much tighter tolerance than 20% -- up to 5%. I'm willing to bet the originals were 20% caps. Just watch out for the inexpensive +80% caps -- nearly double in value for tolerance.
Tantalum caps also typically have +/-20% and can get as good as 5% but not many in the 5% range. You want the + oriented toward the connections to pin 15 and 7. Follow traces on board to figure out which solder pad connects to pins 15 and 7 and use cap's + connection there. C20 is easy to follow - the plus connection is near pin 16 of the IC. C21 - you need to follow connection on bottom of board. Looks like from above drawing that this cap installs in same orientation as C20.
Electrolytic caps -- do NOT cheap out and go this route. These start with a +80%/-20% tolerance and go down with age. Small value electrolytic caps are notorious for being the first caps to go bad in timing circuits and on sound boards.
If *I* were replacing this cap - I would go with a polyester film type capacitor. These are inexpensive and have tight tolerances. A cap such as a Panasonic ECQ-E1105JF would be ideal in this position. Long life, inexpensive, no polarity to worry about and tight tolerance.
Ed
You can also try my system 3 testrom - send me a PM with your email address if you don't have it already and you would like to try it. Before running the testrom you need to disable the watchdog (U11) circuitry. The testrom performs memory and VIA U4 + U5 timer tests and then jumps to the I/O testmode. During the I/O testmode you can enable the watchdog again. If the testprogram keeps running the watchdog circuitry is probably fine, if not the testprogram will stall and stop running.