If it doesn't work, fiddle with it...

'Spinning up' the turbine with an air blast is a pretty effective way to start it, but when it comes to making careful measurements, the fact that something is forcing air into the engine gives a problem.

I can't tell if the compressor is adding anything to the airflow. Certainly the speed increase I get when burning fuel would be accounted for simply by the heat expansion of the air being forced in. (About a factor of three in temperature gives a factor of three in volume - equals a factor of three speed increase)

So: I dug a small electric motor out of the junk box and mounted that to drive the turbine shaft via a rubber band/pulley.
With the engine spinning, I now had only the air being supplied by the compressor and could watch the effect of burning fuel inside the combustion chamber.

When the fuel was ignited there was a distinct and measurable increase in speed as the turbine started to generate it's own power and 'help' the electric motor.

The increase in speed was significant: from 4500 rpm cold, to 7500 rpm at the limit where the NGV begins to glow bright red.

So I'm definitely getting a power increase. Question is: How much?

Observation

It's interesting to note that power is still being supplied after the fuel gas is turned off. The combustion chamber remains hot for some time and continues to heat the air, supplying power to the turbine. Speed gradually returns to the original 'cold' value, but it takes a couple of minutes.

Another motor

Repeated the experiment with a different, more powerful motor. I geared it up so that the motor was heavily loaded and wouldn't max out in terms of revs (that's important - once the motor begins to reach it's own speed limit, internal resistance begins to dominate).

This time I got 7,500rpm cold and 12,000rpm hot. At these speeds the compressor was supplying a lot of air - I could open the gas bottle tap much further before the exhaust reached red-heat. Interesting that the ratio hot speed to cold speed is about the same (1.6 : 1). It's probably telling me something... but I don't know what.

The test self-terminated when the rubber band broke.