(Not correcting CNSparky's VERY good and most to-the-point ever description, but augmenting!)
In a general way, it's possible to say (covering all diesel-electric locomotives ever built here in the US) that the throttle itself either controls engine speed directly and then excitation is controlled automatically indirectly, or else that the throttle controls both engine speed AND excitation directly.
The first instance would be the common EMD locomotives of the 567 engine era (well, most of it, anyway) which have what was in places generally referred to as "LOAD REGULATOR CONTROL." This reference term comes from general Railway Fuel & Operating Officers' Association handbooks and implies that load (or load current, read as amperage) is controlled only by engine speed and any offsets built into the rotating equipment - for example, the common use of a shunt winding carrying a component of load current that acts to oppose the main field at low speed and high current.
What that means generally is that this type of control has very high load currents at low speeds with high throttle settings and overloads or slips without throttle reduction.
The other type of control is more complicated, and originally was developed and championed by ALCO-GE on their road locomotives powered by model 244 diesel engines; this type of load control is called "THROTTLE TRACTIVE EFFORT CONTROL" or else "THROTTLE RESPONSE CONTROL." In locomotives such as this, there are absolute upper limits placed on load current for every throttle notch. The limit is operative at any speed.
What this means generally is that this type of control will have to be near wide open at low speeds to obtain the highest load current, which is limited to safe values.
The first type of control requires no complicated circuitry or anything outside the generator and exciter. The other control needs to be more complex because it needs to be able to place limits on load current for each notch of the throttle. So, then, the first derives excitation strength (field strength) based on engine RPM and locomotive speed only while the second deliberately limits field strength to a maximum limit for every throttle notch.
In almost all cases the governor can prevent the engine itself from being overloaded by lowering (automatically) the setting of the (hydraulically connected) Load Regulator, which reduces excitation. Overload is detected by the governor having to inject more than a specified upper limit amount of fuel for a given engine speed. If that situation occurs, the governor will lower the load regulator setting until the balance of proper fuel rate to engine speed is reestablished.
This is why the first style of control is called "Load Regulator" control since that is usually the ultimate limiting factor. Throttle Response control has many more limits inserted in a more complicated system.
It's much more detailed than that, and complicated too, but what I've given here is a good short way to think of the two widely generalized 'types' of excitation control. You could look at our recent blog article about GE wheel slip control to read about some other artificial limits placed on excitation - in that case, I'm talking about Automatic Power Matching Control. (See the blog link in my signature below the post here.
We actually have some articles planned to cover this sort of topic upcoming on our locomotive blog. First in the works will be a redo of the first primer article that is still visible on the internet archive, on one of our now-down websites. You can see that archive here:
http://web.archive.org/web/200910280543 ... imer1.html We intend to republish it on the new site with only minor changes.
-Will Davis