- he seems to know his shit fairly well, since i'm sure the dyno conversation will come up.
The answers you'll get to this question will depend on what the answer giver uses. I use a mustang dyno so I will explain how that works. The Mustang dyno comes standard with the eddy current controller. This allows the operator to load the rollers for any driving situation and accurately simulate loads based on weight or grade (pulling a trailer up a hill). By entering the weight of a vehicle when tuning you place a load on the motor which is a true representation of what it sees on the street or at the track. Thus tuning is specific for your application. Some other dynos that are inertia only (just spinning the rollers) only load the car as if you engine were in something like a miata. So tuning becomes more of a guessing game as to how the car will react on the street or track.
What's more is the Mustang Dyno uses a closed loop, torque based feedback system. There is a load cell on the dyno which reads the instantaneous torque being applied to the rollers from the car. For one, this is where our hp calculation comes from. HP = (torque x rpm)/5252 and since we know the exact torque being applied to the rollers the math is easy. The torque feedback also allows for accurate real world testing such as 0-60mph and 1/4 mile times. We know the power, weight, and speed of the car. My dyno and proven accuracy of the 1/4 times to be inside a tenth of a second. The machine also allows steady state testing. The steady state can be in terms of rpm or mph. Say I chose 3000 rpm or 40 mph. I can set the machine to hold it at my chosen rpm or mph and regardless of what I do with the throttle the car will not deviate from that parameter. The machine will simply load or unload the rollers to keep it where I want. This is important for tuning and for diagnosing problems that only occur under specific driving conditions. The load cell will also point out small power dips in the tune or drivetrain issues such as clutch/tranny slippage. Inertia only companies also market the idea of their machines being superior in terms of "repeatability." They base this on the fact that you're just spinning rollers and there's no eddy current controller involved. The fact is that I can graph the torque read by the dyno, the voltage sent back to the eddy current, and the load applied by the eddy current. If I wanted to prove repeatability from run to run I could graph it right there in quantifiable figures. So while others say their machine is "repeatable" I'll show you the proof.
There are other companies that offer eddy current controllers on their dynos but don't offer any feedback to the computer so they limit their capabilities. Basically eddy current controllers with no feedback just allow for loading the rollers in terms of percent (1, 2, 3, etc.) and cannot perform many of the functions I just mentioned.
IMO the mustang numbers are the "accurate" numbers. In fact the guy who created the dynojet is on record as saying there's a fudge factor built into the dynojets which inflates numbers because when they designed the first unit it didn't put out the numbers they expected.
That's the quick rundown from my point of view. I'll let others go into the details of other machines.