Firstly, you need to work out what you can happily drive with your engine.
HP, BHP & KW
In terms of work done, one horsepower (brake) equals 746 watts of electrical power. Note that brake horsepower(bhp) is measured power on a dynamometer, while plain horsepower (hp) was an engine rating such as the old RAC method, and bears little relationship to the actual developed power of the engine. When I refer to horsepower in this article I am talking about brake horsepower or BHP. Most engines were eventually rated in BHP, but earlier units were a bit of a minefield, especially horizontal open-crank types.
In practice, engine output is either measured at the output shaft and includes all auxiliaries required to operate the engine, including cooling fan and water pump, or it was measured without these power absorbing bits and pieces, giving an inflated output figure. The Americans used to reckon that British HP were the 'best', followed by the SAE (USA Society of Automotive Engineers) with the feeble Europeans last Some engines such as Crossleys running on producer gas gave considerably more than their nominal rating, which is why it can be deceiving to look at makers figures.
Take a CS 6/1, it develops 6 BHP at 650 RPM. This equates to 6 X 746 = 4476 watts of output power. BUT, you cannot expect to get this out of an engine in actual electrical power, as you have to deduct the (very considerable) losses involved in producing that power. The actual continuous rating for the S-O-M unit is about 2.75 KW. This figure takes into account the losses in the drive method (belts) the losses in the actual alternator, which are winding losses, magnetic losses, slippage (electrical, not belts) connection and brush losses and lastly thermal losses related to winding and magnetic losses. There is an allowance for some short-term overload, but generally you will not get much more from this engine at 650RPM.
A DC dynamo of the same rating will fare even worse, as there are brush frictional losses which are more than slipring losses, and DC machines tend to loose rather more in thermal and electrical losses than AC. If you have an engine such as a small vertical diesel, developing say 4 BHP at 800 rpm (just to be different to the CS example) then you will find that the Lister ratio of input power (in watts) and the output power (in watts or KVA) can be applied with some degree of accuracy, with a downsizing allowance which takes into account the smaller flywheel effect of the smaller engine (taken care of to some extent in the higher rotational speed) Thus 4 BHP = 2984 watts. Multiply by 0.6 = 1790 and again by about 0.85 = 1521. This is a realistic output figure for generated power from 4 BHP. The 0.85 is a factor to cover the smaller engine as already explained. Note that this figure gets smaller as the engine size falls. You may well get more in practice, especially if the engine runs at say 1000-1500 rpm as against the older speeds of 650 etc. So if you find a dynamo, say a Crompton Parkinson 110V at 25A (which I have one of) you can work out the output figure which is 110 X 25 = 2750 watts, then work backwards from the formulae above 2750 X (100 / 95) = 2895, then 2895 X (100 / 60) = 4825. Divide by 746 = 6.46 which brings you back to the 6HP (or thereabouts !) of the CS at 2.75 KW. Note the 0.85 factor has been changed to 0.95 for the larger engine.
kW or kVA
For information, KW or Kilowatt is strictly speaking only used on DC machines. AC machines are rated in KVA or Kilo-Volt-Amps. KVA is only a theoretical rating, as it has to be multiplied by a Constant known as the power factor, written as pf. The power factor can be any figure from unity (one) to less than unity (down to 0.3 or so) Resistive loads such as electric fires and filament lamps are non-reactive to AC supplies, so are rated at a power factor of unity or 1. Inductive loads are reactive, and have low power factors.
I don't want to get too heavily involved in technicalities here, because it becomes heavy going, but if you have a genset with 100KVA output rating, running a load of electric fires, then 100KVA X pf of 1 = 100KW. Inductive loads such as flourescents have a bad power factor of say 0.6, which when multiplied against your original 100KVA, only gives 60KW.
I built some new soundproofed generators some years ago, in which we fitted 50KVA Perkins/Markon sets which we assembled ourselves. Running 10KW lights (240V 10KW each) of a filament type, we got just over the 50KW. Running HMI/CSI lights (these are discharge lights which use a metal halide gas in a quartz envelope, and also a large ballast unit) we only managed to load 30KW. Their power factor was displayed on the instrument panel at just on 0.6. The electricians just couldn't work out why the units couldn't do the same job on the different lights.
If you haven't got lost so far, it gets easier from now on.
Looking out for a machine at sales is fairly easy, but make sure that you have a nameplate/rating plate on it, and the unit is a two-bearing machine (it does not close-couple to the engine, using the engine main bearing as one of the genny bearings) and it turns without nasty noises. The rating plate will give you the operating speed in rpm, and if it is AC then the rating in KVA plus the pf will be displayed. DC machines will be rated in KW, and will have simple volts and amps on the plate.
If you can check the brushes/sliprings, then fine. If not, listen to the unit as you turn it over, any nasties may mean lots of dosh to correct faults when you get home ! The speed rating will be critical, as you do not want an old chugger hooked up to a 3000 rpm two-pole modern alternator.
Two pole machines are invariably 3000 rpm to give 50Hz or 3600 rpm to give 60Hz. Four pole machines are invariably 1500 rpm to give 50Hz or 1800 rpm to give 60Hz.
You will not get into 6 pole or twelve pole stuff, as they start at about 1.5MVA (mega volt amps) and will fill your living room without an engine !
DC stuff is simpler, but the speed range is still critical, both for the engine operating speed and for other reasons. Amongst the finest DC and AC machines are Mawdsleys, based at Dursley and involved with Lister almost from the beginnings of time. These are the best of the available machines and will usually command a high price compared with more modern stuff. Crompton Parkinson were probably as good and in fact existed I think even earlier than Mawdsleys.
A modern machine looks out of place on a 1930's- 1940's engine, so try to match the ages if possible. Try also to get hold of the matching switchboard and control panel if possible, as it will save a lot of headaches later on. Last thing to remember when looking at older machines, don't rotate them backwards, as the brushes can and do object by breaking across the face ! Even older machines with bundles of copper 'brushes' (literally brushes of copper wire strands) will bend under backwards and damage the commutator.
Driving the Generator
Coupling up is relatively easy, remembering that the alignment of the output shaft of the engine and the input shaft of your generator must be very good indeed. Out of alignment shafts will quickly ruin plain bearings on older engines, and will also make the whole set rough running. Belt drives are one solution, but side thrust is something you will have to think about, both for the engine and the generator. Most rotating machines have deep groove ball bearings or better still, roller bearings which can handle a lot of radial thrust, but engines only had plain bearings which also had to take the vertical loading of the crankshaft.
If you want to belt drive, keep the pulley sizes as large as possible, then the belt loads will be reasonable for the transmitted load. Too small a pulley will require too much belt tension which equates to accelerated bearing/journal wear. If you have to gear the engine speed up from say 650 to 1425 rpm, there will be a corresponding loss of torque with the speed multiplication, and while the engine will still develop its original HP, you will loose something in the exchange.
Controlling AC and DC machines is relatively simple, as long as you have the original field resistances and control bits. If you haven't, then it is a bit more involved.
Basically, older machines were classic 'coils rotated in a magnetic field' types, and the output power was collected by the commutator or sliprings on the armature. The limitations of brushes and sliprings limited practical power take-opff, and later on, we changed to rotating field AC machines, where the armature became a large electro-magnet, powered through the sliprings, and the power was collected from the fixed stator windings, which of course had no limit on current capability.
DC machines always stayed with brushes, although modern day small machines tend to be three-phase AC with rectification built in, such as car alternators which give out DC.
The outer field (magnetic field inducing) coils have a current passed through them which creates a magnetic field in which the armature rotates. The field has to be sufficient to induce the required power in the armature. On a modern machine the field power is taken from the output of the machine, and fed back to the field through the voltage controls. This is fine when you start up with no load, as the residual magnetic field in the structure of the machine is sufficient to start a build-up of power which takes a second or so to reach an operating level.
On older machines the same applies, but with lack of use and rough handling, the residual field can be lost. Then you have to re-magnetise the field coils and poles with a battery before you can run. On machines such as my Cub generator, the machine had to be able to start loads with a low power factor, and you cannot do this very well with self excited machines. In this case, the machines have another, smaller generator on top or on the end of the main one, which provides DC excitation current into the field at start up, which is not directly dependant on the output of the generator, and so you can start up with a low power factor or just a heavy load connected. Note that while you can run with under-excited fields, and have a low output, over-excited fields will quickly overheat and get damaged. For this reason, most machines have both variable and fixed limiting resistors in the field circuit, the variable control being used for adjustments under load, the fixed resistor(s) are used to set the upper power limit and also the voltage level off load.