If I understand it correctly, a magnetic resistance unit will give a fixed resistance (or power) per setting, so let's say at resistance setting 3 it requires 200w to turn it. That'll be 200w at 100 rpm or 60 rpm, so probably ideal for fixed or SS, and changing gears on a geared bike will make no difference; 200w (or whatever) is what you get. On the other hand, a fluid trainer will give an increasing resistance as the rear wheel speed increases, so much better for use with a geared bike.
No, not really. With the magnetic resistance units the power is still proportional to speed, it's just that it varies linearly with speed. On fixed, if you are doing 100RPM you are going faster than if you are doing 60RPM and so need more power.
Fluid units have some parabolic curve, so the faster you go the more speed you require proportionally, up to a very high value. That's why magnetic resistance units have different resistance levels, so that they can get the very high power to speed ratio and also the very low power to speed ratio that you get at the start and end of a fluid power curve.
The linear nature of a magnetic unit means that you can use the concept of functional threshold speed (FTS) to build work outs as the FTS is directly proportional to the function threshold power (FTP) and, as the power varies linearly with the speed, 80% of the FTS is equivalent to 80% of the FTP. To do this with a fluid trainer you need to know the curve and read off the values.
Of course, the trainer apps do all this in software and know the speed to power curve. You can use them with either a magnetic unit or a fluid unit, with gears or with fixed. They're not particularly accurate, though, as the system warms up during the workout and changes the relationship slightly.
Fluid trainers have an 'nth order polynomial' curve characteristic.
For electricaly simulated chassis dynamometers, experimental data is acquired by rolling a real vehicle down known gradients or coasting down a level road.
The numbers are crunched to get kW for kmh for 11 points.
Coeffs for a 4th order polynomial are calculated and the control software uses these in real-time through Visual-Basic, DtoA I/O cards and some heavy electrical gear to regulate the resistance on an electric motor.
OTOH, Kurt Kinetic 'Road Machine' is a good enough curve to do serious training on a bicycle.
IME, the 'speed to load curve' is different for every vehicle under the sun.
On a bicycle, a cyclist will increase and reduce the curve by sitting up or going on the drops.
My dyno is set-up to simulate full TT tuck on a Specialised Shiv. I did hill roll-downs to get repeatable kW figures for 32 and 40 kmh. The dyno is 0.75% different at 32 kmh and 1.10 % different at 40 kmh, which is good enough for me.
