Before I do anything to the engine, I want to know how it performs stock, without any modifications. After a bit of searching I found a rolling road that seems to have a good history, Tovami in Goes. They only do rear wheel power, so no stupid calculations to het the crankshaft values.

The rolling road is nothing more than two rollers in the floor where the driven wheels are placed on.

Speedster on the rollers.

To prevent the car from driving of it needs to be secured. With the Speedster this has to be done right to prevent damage to the car. On the fron it was fixated through the front wheels.

Fixated the front through the wheels.

The rear was a little bit more difficult. Normally they just place hooks on the suspension. However, with the Speedster (and the Elise for that matter) it can damage the wishbones. There are two options, either hook it to the rear subframe, or hook it to the whishbones but do it VERY close to the wheels, so the load is at the same place where the wheels put it. BTW, you do need to remove the diffuser to do this.

Fixated the rear at the wishbones near the wheels.

Next, a big fan is placed in front of the car to keep it from overheating and an exhaustgas fan is fitted to the exhaust. The computer is hooked up to the ignition to get an RPM signal and a probe is placed in the exhaust to get the air-fuel mixture. Then the car is ready to roll.

Ready to roll.

After some small calibrations and checks it was time to do the run.

Your browser does not support the video tag.
The run itself

I got a printout of the run and they mailed me a digital version.

\

So the values are:
126HP at around 5500rpm (factory 147HP at 5800rpm)
183NM at around 4000rpm (factory 203NM at 4000rpm)\

So roughtly that would translate to 15% loss, which is quite acceptable. If you use the guidlines from the document on Puma Racing it would translate to around 151HP, which is feasable.

Next day, I did the more interesting experiment. How would my own measurements hold up to the rollingroad run. I wouldn't expect the figures to be exactly the same, but the shape of the graph should similar. The results were very encouraging:

\

So even the vaules compare pretty well to the rollingroad test. The problem is that the rpm values are a bit off, which is due to the way the measurement is done. But now that I know it curve form is the same, I can try and optimize this way of testing. Will be a lot cheaper and much more convienent than to do a rollingroad test after each modification.