I don’t have data on hand to provide you with good answers to your questions. However, we can at least approach the topic using this set of GE main generator curves – I don’t have any comparable EMD data:
Which locomotive model they apply to I don’t know, but given the 1600 hp number and the engine speed schedule, I’d say it was one with an Alco engine with the GE 17MG3 or 17MG6 governor.
To start, if you look at the notch 8 curve, the central concave part, between the two non-differentiable points, represents the nominally constant-power zone where main generator excitation is under load regulator control, intended to ensure that the engine delivers 1600 hp to the generator regardless of what is happening downstream as it were. (And those GE load control systems based upon its own electro-hydraulic governor were reputed to be very precise.)
The two end points of the concave section represent maximum available generator excitation at the high-speed end (upper point) and low-speed end (lower point) respectively. Beyond each of these, the generator operates on its “natural” curve, which as my be seen, and typically of GE practice, is heavily drooped as a result of heavy decompounding. As one moves to the left of the upper point, or to the right of (and below) the lower point, the power drawn from the engine progressively decreases.
Whilst the concave section is nominally a rectangular hyperbola, in fact it deviates. The coordinates for the upper point are 1340 amps and 840 volts (as near as I can read from the graph, anyway), which calculates to 1126 kW, 1509 hp. So that is the main generator power output at that point. The lower point coordinates are 2730 amps and 390 volts, which calculates to 1065 kW, 1427 hp. So the main generator output is lower at the high-current engine of the range. If we assume that the load control system is doing its job properly and keeping the main generator input at 1600 hp, then the power lost to ohmic (I²R) losses in the main generator has increased from 91 hp at the upper point to 173 hp at the lower point, an increase of 82 hp. More current means more power dissipated in the armature winding resistance.
Now if we look at the standstill point, the coordinates are 3130 amps and 100 volts, which calculates to 313 kW, 420 hp. We could expect main generator losses at 3130 amps to be to a bit higher than at 2730 amps, say around 200 hp. So at standstill, the power draw from the engine would be around 620 hp, of which 420 hp is absorbed by the heating of the traction motor windings. In practice, of course, a standstill condition in notch 8, if there were sufficient adhesion and sufficient drawbar load to achieve it, would result in destruction of the motors if maintained for any length of time.
The lower point of the concave curve probably occurs at a road speed of 10 mile/h, give or take, depending upon locomotive design and mission. But the result is that below this speed, any efficiency calculations are fraught with difficulty because engine power utilization is also declining in that zone.
The “natural” generator curve is also load-dependent, as may be seen from the fact that the (dotted line) 4P case is different to the 2S2P case at the high-current end. One may imagine that the 2S3P case would fall somewhere between. It is not inconceivable that the continuous current capability of a 2S3P set of motors would fall well down the bottom part of the lower extension of the resultant notch 8 curve, meaning that the MCS/CTE point occurred where full engine power was not available. This would explain your apparently low-looking calculated MCS power for the low-geared SD-9 case.
I hope that helps – more to follow in due course.