[Jeff Chambers]

It's not that the machine's measurements are traceable to a standard (e.g. NIST), but rather the events under which those measurements are taken. My MD has a certificate of calibration for the load cell and that load cell is 'zeroed' on a regular basis to account for thermal, electrical and mechanical drift. But let's look at the physics involved. HP is a measurable quantity, however its a dynamic quantity. 1 HP = 550 ft-lb/sec. These dynamometers (and their computers) measure a reaction force on the roller support (or PAU support) at a known, fixed lever distance so they can directly measure a STATIC torque (ft-lbs). That static torque doesn't directly correlate into work until it is moved through some distance over a period of time. Now given a known mass (inertia) of the rollers and some measured acceleration rate, work can be calculated (inferred). These are not simple calculations, for example, the rate of acceleration has to be calculated from finite velocity measurements taken from the rollers. Acceleration is the first time derivative of velocity (calculus) so its calculated accuracy is dependent on how well roller speed can be measured, the sample frequency, filtering of the signal noise from the device, etc. The way that each dyno chooses to do this affects the outcome. They may be measuring the same thing (the load eventually resolved at the load cell), but they are inferring (calculating) a completely different measurement from that load.

Now let's talk a little about what outside of the physical dyno affects the measurement and try to understand why the end user should look at the results as somewhat qualitative and not purely quantitative. Here are just a few parameters that affect not only the power the car is making, but also the power that the dyno is trying to calculate:

humidity
barometric pressure
temperature
engine temperature
transmission temperature
rear end temperature
oil temperature
fuel pressure
tire grip (at any instant throughout the run)
electrical line noise
rpm pickup (inductive, capacitive or optical)
rear gearing (final drive ratio)
rate of acceleration (yet another derivative)
phase shift and lag (of all signals)

These are just a few of the many thousands of variables (imagine the inputs and variables in your PCM alone) that I can drum up on this sleepy Sunday morning. So just as Ford and GM can start with the same exact 305 cubic inch displacement and wind up with completely different power, fuel economy and performance, the different dyno makers are going to wind up with different measurement systems.

Users also need to remember that they directly affect the outcome. As an operator, I get customers that don't want to spend anymore than the $65 for two basic pulls. They come in unprepared, but expect the world. Next time you're at the dyno, take copius notes then go back home and ask yourself just how accurate the results really are when you've run your car with the innacurate idiot gauges in the dash and no idea how many of the variables compared from run #1 to run #2.

Like I said, the idea is to strive for consistency in some form and fashion so that you can REASONABLY assess how your car is performing. This isn't rocket science, if it were, it'd cost you alot more than $65 for a couple of pulls.