first something for J.Guy :

optocouplers are useless if you have an emi problem with a high frequency peak that has enough power. the emi just "jumps" over the isolation and you have the peak on the other side of course not with that power but the peak is there. optocouplers do galvanicaly seperate two circuits powered from different sources but they are not "that" what is needed against emi but they do also help (a bit).

one point : if you can kill the emi noise where it is generated before it can spread around in your circuit. the closer to the emi source the better it is.

next emi is energie that needs to be converted in heat to be removed so that is the trick of ferrits. if you know the frequeny(range) of your disturbing signal than you can calculate the needed cap(s) to filter that signal in your signal. again, right calculated (or luckyly choosen types) have a low impedance for that cirtain frequency or the frequencyrange and again, that noise is turned into heat. yes i have seen caps getting hot because the energie of the frequency was to high. that is the reason why you have a frequency derating in the cap datasheet for the current (power) that cap can handle.

one very important point is a propper layout and placement. you can add millions of parts if your placement if your layout is not considered with emi thoughts. and coupper is there during manufacturing of your pcb and you have paid for that area, so use it. in most cases you do not need additional parts if your layout is setup correctly.

but back to roger. you must have a brother :-) he has the same name, same hobbie and ... believe it .. he is doing the same job :-) fox mulder would say "i want to believe it"

roger i do not know your hardware how it is setup and the signals how they look like.
you say you have an ignition system so my first guess would be that you see the ignition pulse or something that is synchron to that. if that is true then the source is the ignition coil. so start from there. how does your circuit look like ? i would guess you have the coil directly connected from the driver output and against gnd. anything else ? if not you should ! if you open the switch to switch the ignition coil off, it will generate a high peak pulse that goes back to your driver circuit, well that circuit is protected but the pulse is in this circuit and spreads around.
for such high peak pulses that are very short you should add a tranzorb (this handles fast transients) in parallel to your driver output, but as close to the coil as possible, again the closer to the noise the better. in addition to that add a varistor that part can take a lot of power, but is slow compared to the tranzorb.
next mount some ferrit beads over the cable to your coil these will damp the noise very effectively.

if you use caps to filter something be aware that the cap has a second connection (gnd in your case) and that this connection needs to be - from an emi point of view - a low impedance connection otherwise your filter cap wont work very well. where should the emi noise flow if it sees an high impedance ?

your switch from infineon has a power suply pin, have you made any filtering there ? your noise can also flow from the source over the infineon switch to the power supply and from there being spreaded around.

se i would look for that noise and where it is generated. then see how it can be avoided to spread around. i cannot give you exact values for the parts (well i can help you with the varistor and the tranzorb if you give me more information like voltage, current f.e.) because this is too application depended.
but maybe we can do some netmeeting and you show me your hardware and the signals and i get a better impression what you could do. but this weekend i will be offline except sunday evening german time zone.

keep me informed.

regards.

Michael